Hey HN, I've been building Sectr, a backend-as-a-service for making it easy(ier) to develop and deploy AI agents. Deploying agents requires non-trivial systems design work. Consider a support agent: how do you stream messages to your frontend, how/where do you persist messages for frontend fetching later (or for auditing), and when the agent wants to issue a refund, do you spin-down the compute infra while it's waiting on human approval? Sectr aims to be the serverless abstraction and platform to provide these systems for you. No sign-up needed to try using the dev server: curl -fsSL https://sectr.dev/install.sh | sh # Install CLI sectr init my-agent && cd my-agent # Starter code sectr dev # Dev server, with UI to test chatting Sectr provides: - Streaming of every token, tool call and event, and persistence so they can be loaded fresh later. - Scaling up, and down to zero when the agent needs to wait, and resumption right to the suspended agent's state. - Human approved tools automatically use Sectr's suspension and resumption capabilities, no code needed from you. - AG-UI endpoint for any deployed agent so you can trivially hook up your frontend - Local dev server experience, so you can rapidly debug and iterate on your agent's behavior At the core, you can deploy any container image to Sectr as long as it implements a certain shaped HTTP endpoint. In practice, currently Sectr provides adapters to let you deploy your existing OpenAI Agents SDK, LangGraph, or Pydantic AI agents, with only 2 new lines of code: app = AgentApp(name="support") app.add_route("/chat", openai_agents(agent)) Certainly the hardest part of building Sectr was getting the resumption logic to be correct and invisible, across the Program State x Model Provider x Agent Framework matrix. For example, an earlier bug was when a model replied with text plus a tool call, Sectr would resume that as two separate messages, instead of one message with a text and tool part. This works, but breaks prompt caching (and might subtly change model behavior). Sectr is free during the beta (and can be tried locally without sign-up); I mainly want feedback, especially on dev & deployment experiences, any bugs, and what features would excite you. Introductory video: https://www.youtube.com/watch?v=H-y3wynkKc0 Docs: https://docs.sectr.dev
In an RTS game, making decisions quickly matters as much as making good ones. Looking for the balance between cost, speed and intelligence in agents, I started wondering what an agent playing an RTS would look like. So I built a simple RTS, and I keep making it more complex. - The game keeps running while the agent thinks - The agent can upload a JavaScript script for each unit type - 300+ game replays - Connects through MCP or CLI. Works with Claude Code, Codex, Cursor CLI, Antigravity CLI, OpenCode, OpenClaw, DeepSeek Harness and MiniMax Code - 2 to 6 players, FFA and co-op, open and walled maps - 4 units, the King and 4 buildings - Open source, written in TypeScript
A WooCommerce plugin that lets AI assistants like Claude and ChatGPT search your products, check live stock and prices, and send the shopper a pay link on your own checkout. Browse-only by default. No marketplace cut. Orders land in WooCommerce as usual.
On Record Verify lets any AI assistant or platform check whether a photo or video is a genuine On Record capture: when and where it was taken, and whether it has been changed since. Paste the code on the picture into Claude or any MCP-capable assistant and it answers from the record, not a guess. Every capture is sealed at the moment it is taken by On Record's Authenticity Chain and anchored into Bitcoin. The proof is public, so nobody has to take our word for it. Free, read-only, no sign-in.
Thinker cuts your coding agents’ token spend by ~20%. Thinker is an open-source proxy that sits between your agents and the model. It strips context your agent no longer needs, caches knowledge so future agents don’t have to rediscover it, and keeps your instruction files lean and current. Across my first dozen users, it’s averaged 20% fewer tokens per completed task, with no drop in success rate. For teams, it also shows where token spend is going, by team, repo, and agent.
Hello, my name is Berend Graham. My cofounder Shiv Oberai and I have been working on Limma for a year now with the goal of building a proactive solution to the reliability issues that plague practical AI and agentic use in companies with RAG. Limma is a single MCP that exists for the sole purpose of ensuring that the right information is retrieved with every query and no information is skipped and or missed. This means total precision and recall. We've found that this approach leads to the lowest rate of hallucination vs. solely targeting agent action verification which has made tremendous strides in capability. We do this by building a map and structuring all data-points within unstructured data as well as federating structured data for addressability so that agents don't need to search through chunks directly and instead only get the right information . We wanted this to be extremely easy to use so that anyone can set it up and try it within minutes so that the precision issues (especially with large quantities of data) that LLMs struggle with can finally be remedied. We use self-hosted models (primarily Qwen models as of current) so that the system can be deployed on one's own server if needed, and also to optimize for speed. We have a free monthly version on our website if anyone would like to try it out and provide feedback!
GoZapp is a job search app for France. Its free, unlimited AI Agent reads your profile (skills, experience, location, salary), scores every new job offer out of 100 and sends you the best matches. GoZapp also writes your CV and cover letters with AI and checks your CV against applicant tracking systems (ATS). On the web, each tool costs 1 euro with no account. Available on iOS, Android and the web, in French.
An approachable, AI-native image editor: layers, real photo adjustments, retouching, artistic effects you can paint with, on-device AI, asking for edits in plain words, and an MCP server for AI agents. Not a PhotoShop clone! Free in a browser, no account needed. Use brush with every effect!
Token Kingdom turns your Claude Code usage into a cozy 3D village. Every token you spend becomes gold, wood and stone. Build, fish, cook, defend against raids and visit your friends' kingdoms. One plugin install. Free.
Since our first launch, GS Copilot has grown into an AI coding agent inside the Apps Script editor. New: Agent Mode for multi-file changes, Plan Mode, "Fix with GS Copilot" to debug errors in one click, inline suggestions, and MCP Tools on Pro and Lifetime. You approve anything that sends, deletes or touches a secret. Launch offers until Oct 31: code PRODUCTHUNT takes $30 off Lifetime + BYOK ($99 one-time), and code PHSUB takes 25% off monthly Plus or Pro for your first 3 months.
Doska is a whiteboard + docs where diagrams are code. Write Mermaid or D2, or drop terraform show -json, and get editable shapes synced with the code. Embed live diagrams in docs, let Claude edit via MCP. Free sandbox, no sign-up.
Raq.com started as an internal tool builder. It's now a toolkit for the AI agent you already use. Connect Claude, ChatGPT or any MCP agent once and it works across 75+ business tools, from CRM and task boards to quotes with e-sign and card deposits, forms, automations and AI image and video. Everything it makes is saved in one workspace your team shares, and you can build and host your own custom tools too.
Free desktop app for API work on localhost: *.localhost proxying with capture, a client for REST/GraphQL/gRPC/SOAP/WebSocket/MCP & more, proxy-level mocks, git-synced workspaces, visual flows, a local mail inbox, and an agent-ready CLI. Windows, macOS, Linux.
New since our June launch: SecondBrain is rebuilt around your voice, and it now runs in your browser too. Tap and talk: it writes a clean note in your words, files it, and turns dates into events and reminders. No time given? It asks. Share links, PDFs and photos from any app, then watch it work live. Ask anything and the answer streams from your notes. Pro adds one shared memory for Claude, ChatGPT, Gemini and Cursor over MCP. iPhone, Android and web. Free: 10 notes and 3 AI requests a day.
Generate 3D, video, PBR materials, audio and images without leaving Blender. Scenario for Blender is an open-source extension (GPL-3.0, 800+ GitHub stars) that imports each result straight into your open scene: a photo becomes a textured mesh at your 3D cursor, a move you describe plays on your rigged character, one line of text blocks out a room, and your camera move comes back as a photoreal shot. One click connects Claude Code, Codex or Cursor too.
FYN is personal finance with one ledger for you and your AI tools. Link banks once on Pro. FYN categorizes transactions, keeps budgets and bills current, and turns receipts into line items. Ask FYN in-app, or connect Claude, ChatGPT, Grok Bot or any MCP client for a read-only view. FYN never moves money. Free forever: receipts, tap-to-pay and email ingest, Ask FYN. Pro: Plaid bank sync, higher limits. Households share one workspace. iOS, Android, web.
Velocity tracks live aircraft (ADS-B) and ships (AIS) from public feeds with no API keys, records every position it sees to your own disk, and lets you drag a scrubber to rewind the map. Aircraft also replay from public archives back to 2024. Area alerts go to Discord, a webhook or ntfy. Every contact shows which feed reported it and how old the fix is. It also runs as an MCP server, so an agent can ask what is flying somewhere right now. docker compose up, AGPL-3.0.
Create AI product videos from photos or a public product link. Connect Imaginou to an assistant over MCP and OAuth, review the scene storyboard before generation, then inspect and download the finished video. Choose showroom, presenter or comedic campaign workflows in four languages. Three free trial orders are available once per verified account, followed by one-time order packs.
Describe the job in plain words and Whenever's AI builds the workflow, asking about your business rules, not technical settings. AI writes it once, then code runs it. Add AI steps or agent steps only where the job needs judgment. 1,400+ integrations plus any API. Runs on schedules, webhooks, app events or on demand; every run and step logged. Build in the app or from Claude Code, Cursor or Codex over MCP. Each workflow is a TypeScript file you can export. Free plan; Pro $39/mo, priced per run.
Scento Perfume API is a remote MCP server and REST API. Paste into Claude, ChatGPT, Cursor or VS Code and twelve read-only tools appear. It covers 1,400+ designer and niche perfumes and 400+ raw materials, and needs no key or account. With the free browser extension, a small card on a Fragrantica perfume page shows what a 2, 5 or 8 ml sample costs, its rating and its leading notes.
AgentBreakWatch tracks release velocity, breaking changes, and MCP tools/list schema drift across 258 production AI Agent SDKs and Model Context Protocol servers on npm and PyPI. Paste your package.json, requirements.txt, pyproject.toml, ormcp.json to spot archived servers and breaking upgrades in 5 seconds, or query our free MCP server directly from Cursor and Claude Code.
Asonix helps iOS app teams find keywords they can actually rank for, compare opportunities across 175 countries, and track rankings daily. See competitors' Apple Search Ads and Meta ads, monitor AI visibility, and connect to AI agents via MCP.
Filmit.io is a suite of tools for editors and producers. • 50+ plugins, effects and transitions inside Premiere Pro, After Effects and DaVinci Resolve • Filmit MCP: connect Claude to Premiere, After Effects and Resolve to cut, add markers and build graphics • Filmit Studio online (beta): plan your story anywhere, review with clients, send it to Premiere, plus a production suite for estimates, schedules, call sheets, crew and releases One subscription, 7-day free trial.
Taxley lets directors of small UK limited companies prepare and file their own Corporation Tax return and accounts online. Free to prepare; pay per filing, no subscription. Free MCP tools for AI assistants (Corporation Tax, deadline, penalty, director's loan and capital allowances calculators) at
WP AI Manager is an AI agent inside your WordPress admin. Ask in plain language and it edits Elementor and block editor pages, writes SEO, runs WooCommerce and 12 playbooks. Approve big changes, undo any. BYOK MCP for Claude and Cursor & more.
An engine that learns your coding style from past projects, turns it into agent skills, and proves coding agents follow it. Claude Code today; Codex and other agents next.
Create a JSON approval policy for risky AI agent actions. Pick action, environment, sensitivity, impact and reversibility. Runs locally with no account. Optional $19 pack adds editable policies, reviewer runbooks and integration checklists.
yovoice is an open-source, local-first studio for turning words into expressive speech, cloned voices, full songs, and multi-track scores. Run everything on macOS, Windows, or Linux with no cloud API calls or per-character fees. Build stories and music, compare takes, edit timelines, export stems and MIDI, and connect AI agents through the built-in CLI, HTTP API, and MCP server.
Search and compare 322 open-source AI and vibe coding tools: chat UIs, model servers, RAG, agents, coding agents, Claude Code skills, starters. Each gets a score from public GitHub data, a review and its license in plain words. Nobody pays for a spot.
FlutterGo.AI is an AI Flutter app builder that ships real apps to both stores. Describe your app, or start from Figma, photos, a ZIP, a GitHub or a URL, and get a working Flutter/Dart project with a live preview. Edit the code, connect Firebase or Supabase, sync to GitHub, and build signed releases for Google Play and the App Store. A Flutter MCP server lets ChatGPT, Claude, Grok or Cursor build apps in your account.
Since the first launch: weekly timesheets built from the clock with overtime and one-click approval, a shift rota, hourly pay from attendance hours, automations, and an API with webhooks. Tasks, automatic time tracking, attendance, leave, payroll and HR sit on one employee record, with an AI agent that shows its plan and waits for you to confirm. Free for up to 5 people, then $8 per member per month.
Jean is a free bioinformatics and biotech suite for the whole experiment. Design constructs, primers and CRISPR guides; simulate digests, PCR, Gibson and Golden Gate builds; analyze Sanger traces, qPCR and ddPCR runs, variants, trees, mass-spec proteomics and molecules; verify the clone you got. Every engine is checked against its field's reference tools and every result explains its method. It runs in the browser with no account, and as a remote MCP server any AI assistant can call.
Pruneo frees up space on your Mac: caches, app leftovers, old versions, large files and the dev junk other cleaners skip. New in v1.1.0: it adapts to the apps you use, asks your AI agent about unknown folders and has an MCP server. Free, with undo up to 30 days.
TenseMem is durable agent memory with a deterministic resolver, enforced cardinality, freshness-aware reads, and lasting corrections — so your agent gets what still holds, not a pile of conflicting prefs.
Agent Index is a free, sourced directory of about 404 AI agents for compliance, security, risk and business operations, plus about 483 independently verified audit firms (SOC 2, ISO 27001, PCI DSS, HITRUST, CERT-In and more). Compare tools side by side, filter by framework, region and integration, and find auditors by country and city. Not a ranking.
For AI builders who fear bloat, bugs, and broken intent. SPIDER Protocol turns your AI coding agents into a governed engineering team: • Preflight Discovery: Absorbs true user intent before touching code • Living Git Memory: Architecture & decisions immune to context amnesia • Independent Validator Gate: Audits the diff and full test suite before any task is marked Done Build production-grade software at AI speed. 100% free & local.
I originally built drop.media to do one thing: drag an image into your browser and get a shareable link. It still does that! Now it's one media library for you and every AI agent you use. Upload from the browser, terminal or API, or let Claude, ChatGPT and Cursor save files to it through MCP. Any agent can find what another saved, use it, and share it as a public link or one that expires. Images, PDFs, video and docs, on every plan, including free. Upload once. Use it everywhere.
Kinaya is a keyboard-first browser for macOS. Ads and trackers are blocked inside the engine (Brave's adblock-rust with EasyList and EasyPrivacy) before a request leaves, and there is no telemetry. A built-in MCP server lets Claude Code read pages, click, type and take in its own background tab, using your real logins. Also: command palette, private and Tor tabs, encrypted password vault, reader mode. $19 once, 14-day free trial.
GetUrUI gives you 106 free, MIT-licensed animated React components, blocks, charts and AI-chat UI, built with Tailwind CSS and Motion. Install any piece with one shadcn CLI command and own the source. An MCP server lets Cursor, Claude Code and other AI agents find and add components for you. Every item ships in light and dark. Pro unlocks all 332 designs, and 32 full templates cover SaaS, dashboards, commerce, docs and wellness businesses.
Upload a video or audio file and get translated subtitles (SRT, VTT, ASS), a voice-over track and a finished video in 35 languages. One clear AI voice reads the translation while the original sound dips under each line, like a TV documentary. Pick from 179 voices, add SDH or burned-in subtitles, and see the price before you start. The engine is open source (AGPL-3.0), with a CLI and an MCP server for AI assistants. New accounts get 60 free credits.
Knot Platform is the open-source stack for a Knot: a node where people and personal agents share the same records. Agents can post what they need and what they can offer, keep context private or shared, and leave results humans can read—on a node you self-host. I designed the exchange architecture and open-sourced it. For anyone who wants to run their own agent-era information hub. Live:
Clean your email list before you send. Upload a CSV or use the API: every address comes back valid, invalid, catch-all, risky or unknown, with a score and a plain-English reason. What's different: catch-all and unknown results are refunded, duplicates are charged once, credits never expire, no subscription. 100 free checks every month, EU servers, API, webhooks and an MCP server for AI agents.
Wingo UI is a React 19 + Tailwind CSS v4 library with 226 components and 95 blocks that install as source code, shadcn style. Every component is animated, configurable through props, works in light and dark, and is built mobile first: bottom sheets on phones, 44px touch targets, swipe actions with a visible fallback. A CLI adds items with their dependencies, and an MCP server lets Claude Code, Cursor and other agents search the library and follow its design rules. Many items are free.
OpenRadView opens the DICOM and NIfTI files from your scan CD or zip on your own computer: linked axial, coronal and sagittal views, experimental 3D and measurements. Optional MCP access for your own AI agent. It is a personal research project for education and has not been clinically validated. Ask a qualified clinician to interpret your scans; do not use the app for decisions about care.
The ElevenLabs Realtime skill teaches AI coding assistants to build on ElevenLabs Realtime models. It sets up an ElevenAgents agent on Realtime Mini or Max, helps pick between them, and migrates a browser voice app from OpenAI’s Realtime API by swapping the credential, WebSocket URL, model and session config. MIT licensed, part of the elevenlabs/skills repo, installed with npx skills add elevenlabs/skills.
ondocs adds an Ask AI box to your docs: visitors get a short answer from your pages with citations to the exact section, or "I don't know" when the docs don't cover it, on your own model key. Coding agents get an MCP server (search, fetch, list_pages) and llms.txt from the same index. Plugins for Starlight, Docusaurus and VitePress, guides for Next.js, Fumadocs and Nextra, or one script tag. No vendor, no vector database. MIT licensed.
The open discovery directory where indie software, developer utilities, and AI agents gain lasting visibility. Ranked by launch freshness and real usage, so great tools never get buried in 24 hours. Free submission, dofollow backlink, and AI agent discovery via MCP — queryable inside Cursor, Claude, and Windsurf.
Waspes 2.0 adds a 3D Studio that generates textured 3D models (.glb/.obj/.fbx) from text or images and streams them live into Roblox Studio, plus a public remote MCP server ( ) with 8 tools so AI agents can build websites and 3D assets. Still includes the AI website builder: one prompt gives a production-ready HTML5 site. Plans from $2.50/mo.
Graphicle is a Rust port of Graphify's open-source code-graph engine, 3 to 22.7x faster than Graphify 0.9.91 in all 39 of our measurements, plus project memory: docs, decisions and past agent sessions, cited to file and line over MCP. Local, Apache-2.0.
Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, security risks, prompt injection, data exfiltration, and supply-chain risks in Claude Code, Codex, and MCP skills before you install them.
All-in-one chemistry MCP server for pharma and biotech: over 250 tools in 18 families for spectra, stability and toxicology. Try it free in Claude, ChatGPT, Cursor. 14 to 41% without CovaSyn, 76 to 92% with CovaSyn. Measured on MolecularIQ, a benchmark we did not design (arXiv:2601.15279, Klambauer Lab, Institute for Machine Learning, JKU Linz).
A desktop Git client for you and your AI agent. You get a fast graph, line staging, drag-to-rebase and Undo for Git. Claude Code, Codex or Cursor connect over local MCP: they see what you selected, use far fewer tokens than raw git, and only propose commits.
Your whole team, one flat price. Pro is $35 a month for the whole workspace: no per-seat fees, unlimited people and documents. Send a document, get it signed on any phone, and keep one sealed PDF with a certificate. Signers can pay while they sign, straight into your own Stripe account, and Pro adds no fee. Build on it with a REST API, five SDKs, a CLI and an MCP server, or run it on your own server for free (open source, AGPL).
LemMeBuyIt is the retail data layer for AI agents, resellers and developers: live prices, 12-month price history, cross-retailer compare, ASIN and Buy Box mapping, and clearance signals across 100+ U.S. retailers. Free MCP server, one-click sign-in.
Open-source MCP server for Claude Desktop and Claude Code. Connect any number of Gmail and IMAP accounts, then ask what's unread everywhere, who's waiting on a reply, or to draft answers. Runs locally, logins encrypted, drafts-only by default.
Veynod is a shared blackboard for multi-agent teams over MCP and REST. One versioned memory, task hand-offs with leases, independent quality gates, human approval, and delta context sync for Claude Code, Cursor, and custom agents.
KnightCode is an open-source coding agent for the terminal. Sign in with Claude, ChatGPT, Grok or Copilot, use an API key for 40+ providers, or run a local model. Every file read, edit and command shows up in the transcript. Sessions are a tree you can branch, fork and compact. Extensions, skills and themes live next to your repo. The system prompt is about 1,100 tokens, so more of the context window goes to your code. MIT licensed.
EngMock turns your own interview questions, interview write-ups and notes into live AI mock interviews for software engineers. Practice system design, coding/DSA, OOD and AI-assisted coding with adaptive voice follow-ups, a whiteboard or coding workspace, and scored reports with evidence and replays. VS Code mode (preview) keeps coding in your editor and voice in the browser. Use MCP to create and edit scenarios, prepare interview links, and review reports and progress.
A Chrome extension for developer documentation. Open a developer docs site, click scoop.md, and its llms.txt plus its instructions for agents are copied as a short prompt for Claude Code, Codex or Cursor. No llms.txt? scoop.md builds one. Free, open source.
Red-XAI is a growing technology ecosystem by Red-XAI LLC, featuring Vortex's free AI chats and autonomous agents, Red-XAI Host for self-hosted websites and apps, browser-based Cloud PCs, and Red-XAI DB with its own query language. Build and integrate with APIs, SDKs, and CLI tools. Red-XAI Hub is coming soon. Investors, strategic partners, businesses, and app integrations are welcome. Explore
StartupKit is an MCP server and CLI that gives AI agents tools to research a market, launch a product, and grow a startup: a free production-ready SaaS stack, SEO and keyword research, and launch checklists.
Hey folks! I made it easier for builders to use the lebanese legislation through: a public DB, API, and MCP. The Lebanese University has an archive of all laws dating back to 1904 (legallaw.ul.edu.lb). People and AI can barely find it and navigate it. It also had outdated articles. Now, more people are using it. Builders for their startups, lawyers for their work, and students for their bar prep. I'm still trying to get it trusted by authorities. The database includes the edges between every law, article, ruling and Gazette issue. You can easily identify how they influence one another. Articles are versioned, so you can see how it was affected by other laws from the date it was published. 100k unofficial english translations were added using LLMs. They aren't accurate, but they help with search and SEO/AEO/GEO. We're working on producing legally accurate translations. 47k rulings were PDFs, some handwritten. We converted all of them to text using OCR, flagging handwritten files. We're working on an HTR model specifically for those to increase the confidence. Today, other datasets in Lebanon charge for this data. I'm doing it as a public service. I want change in my country. And, lately, I think it's through helping builders. Yet, what I'm doing is legally wrong, but hopefully morally acceptable.
Paste a text or upload a file and see how likely it is that ChatGPT, Gemini or Claude wrote it, with a score for every part. Most detectors are trained on English; carrierscan has its own detectors for Ukrainian, Kazakh, Uzbek, Estonian, Georgian and Lithuanian. On the same Ukrainian texts it caught 16 of 17 AI texts (GPTZero: 14 of 17) and flagged no human ones. Free, no sign-up, texts never stored. Also a public MCP server for Claude, ChatGPT and Cursor.
This is hardly a greenfield area. There are tons of agent-to-agent communication apps. But, at least to my taste, a lot of them are too complicated or too opinionated. If we learned anything from the HuggingFace incident, it's that modern AI agents don't need a lot of help communicating. They are smart and creative and figure it out. So...with that in mind, I (meaning I + claude and friends) built kashkeshet, or kkt. It's a small, Rust-based MCP server that gives agents a few simple chat primitives (identity, channels, chat) and leaves it up to them to use those however they want. You as the human get a little web interface so you can watch the fun. Coordinate your coding agents; setup a virtual company; conduct sociological experiments; whatever. kkt's main goal is to stay out of the way, and let you and your agents figure out how to organize and communicate on your own. This is meant to be run on your local device and shared among local agents. There's a little explanation of how you could use tailscale to go broader, but nothing like Oauth2 or anything yet.
Billy is a free, open-source invoice editor that feels like writing a document. Click any line and start typing, let totals calculate themselves, and export professional, print-ready PDFs in seconds. With MCP support, Billy can also integrate with AI-powered workflows. Built for freelancers, indie hackers, and small businesses who want simple invoicing without the bloat. Try it at
US federal contracts, defense contract announcements, lobbying, congressional and insider trades, tied to the ticker that was valid on the day each record became public. Delivered as Parquet files, a REST API and an MCP server for AI assistants.
Implemented agent smith since start 2026 with over 1000 specifications which lead to about 500k lines of code. The procedure with that spec-first works well. How do you do it?
Cryo is a programming language that models long-running tasks like workflows or AI agents. Programs freeze to disk when idle and can be thawed when new data arrives. As part of our announcement of Endothermic Labs, we've put together a small of Cryo that highlights its features and walks through the language and its plugin system. We originally built it as a way to deploy healthcare agents at scale, but have rebuilt it from the ground up to be a standalone toolchain for long-running tasks. It acts as something like a cross between Durable Objects and Temporal workflows, where the script itself is the durable unit with no replay needed. Cryo is statically and structurally typed, with a concurrency model similar to Go. It's still under active development, so we're holding off on open sourcing it until we're confident that the language is stable, ideally in the next 3-4 weeks.
Hi HN, I'm Alexander. Yes, this is another agent harness. Here is why I built one anyway. I was engineer #2 at Browser Use, where I built our agent harness and the evals behind it, including BU Bench, which became a default benchmark for browser agents. I left in September to build Mainplane. The problem: local agents die with your laptop, and cloud agents can't reach your work. Many cloud platforms ask you to "select a repository", but my work is not that simple. I work across a desktop, a Mac mini and a laptop, and much of it is not in git. So I loaded Ubuntu on my old 2012 desktop and made it my Mainplane server. It has run for four weeks at a load average of 0.02, because the agents' tools run on other machines. How it works: it is a control plane architecture. The server runs the harness. It holds your API keys and every session, makes the LLM calls, and routes each tool call to a worker. The workers are your other machines, connected over a built-in private WireGuard network. A bash command runs in the worker's terminal, not on the server. Shared drives let projects move across machines. The agent works on your machines just like you would, but its state lives on the server, and your keys never leave it. Some design choices: - Every session is one append-only file. This is the agent state, and it is easily readable. If something breaks, you can resume it from the file. Observability comes for free: agent traces are just these state files. - Small by design. The system prompt is 3k tokens, and there are just two kinds of tools: shell and file (read, write, edit). At Browser Use, I learned the bitter lesson and found frontier models did better when I deleted all our helper functions. So I made the same bet here: bash is all you need. - No compaction. When a session fills up, /continue starts a new one that greps the previous state file, retrieving the context it needs, instead of reading a lossy summary. A /compact on a 400k-token session takes a while to write its summary. A /continue into a brand new session starts at once. - No permission prompts. A hyperintelligent misaligned agent WILL find a way around a permission system. The better strategy is giving the agent full unbounded access to a local computer, and only putting what you trust the agent with on that computer. - Why Go. WireGuard is written in Go, so the network compiles right into a 20 MB binary. Go is made for production servers, LLMs write it well, and the simplicity of Go matches how I like to build. Mainplane works with a Claude or ChatGPT subscription on the regular usage limits, or any API key. I would love your feedback, especially brutal and honest feedback and tasks that the harness cannot do. I'll be here all day to answer questions.
Turn your Claude Code and Codex sessions into a living fantasy village. See agent activity at a glance, then click a hero to inspect context usage, skills, plugins and subagents. Agent Quest reads existing local session logs while you keep using your usual tools. Keep multiple sessions in sight on a second screen. Customize the village and export Markdown or JSON reports for review. Free, MIT-licensed and open source. Runs locally in your browser with Bun.
Coding agents write blind. They can't run your app, see the screen, or check their work — so you end up running builds, pasting screenshots, and copying console errors. Vibex is a native Mac window that gives your agent eyes. It embeds a live iOS simulator, browser preview, source control, and AI usage dashboard with 50+ MCP tools. Claude writes code, runs the app, taps the simulator, reads the console, and checks every diff. You just review. Free forever. MIT licensed. No account, no tracking
Your agent is thinking, calling tools, waiting for an answer, or stuck. Agent Mascots shows it at a glance: 23 animated characters, 8 states, any color. One dependency-free web component, plus an agent skill that wires it up for you.
Stop prompt injection, secret leaks and risky MCP tool calls before they leave your AI agents. Works with Cursor, Claude Code, VS Code and GitHub. Free for 5,000 requests a month.
New in 1.0: decision models next to LLMs. They answer closed questions with the probability of each answer, fast and at a low cost, and a model group sends each question to the model that suits it. Also new: Docling for your files, and tools for agents.
Hey HN, I'm Christian and this is a new OSS project I've been noodling on for the last couple of weeks. Hob is a free, MIT-licensed desktop app for Mac, Windows and Linux that works on your files the way the Claude and ChatGPT desktop apps do, except the model runs on your own computer. It uses open-weight models (currently Gemma 4 and gpt-oss). All your chats and files never leave your system. The only things that go online are web search (can be turned off in settings) and any MCP connectors you add yourself. Main features: - Automatically finds the best open-weight model for your hardware on first install (Mac/Linux/Windows) - Writes sandboxed Python code for ad hoc tools (data processing, document conversion, graphs etc.) - Built-in web search (powered by SearXNG) - MCP support - Image recognition - CRUD to your local filesystem - Works fully offline with no internet Before building this I tried every local AI app I could find, but it seemed each one had different problems. The main thing I noticed is that none of them could do what the ChatGPT and Claude apps do with your files, e.g. read and change files in a folder, look at images, run code in a sandbox, draw a chart. The most advanced thing any of them did was paste a plain text file into a small context window, which basically failed as soon as you had a file larger than a few KB. I couldn't find anything that could easily parse images and bigger PDFs. The OSS solutions that were more advanced all seemed to default to hosted models (either OSS or closed), which defeats the entire point of the privacy that on-device OSS offers. They also seemed to push towards an account/open-core style business. I build it using Electron bridged with llama.cpp in its own process. On initial load, the system will try and figure out if you have a compatible graphics card and if not default to CPU-only (I ran into a lot of issues of graphics card beeing seemingly compatible but then crashing on first query, there's a whole saga to this lol). Code execution happens in a sandbox that can only write to the chat's folder and has no network unless asked (sandbox-exec on macOS, Landlock on Linux and a low-integrity process on Windows). Images go to a small Qwen3-VL model. Dictation is a local Whisper model. The hardest part was definitely to get the smaller models to use tools reliably. Ran into a ton of issues of 4B models happily saying "here's your chart" without ever calling the chart tool or get stuck in an infinite call loop. I spent a fair amount of the work building guards vs tweaking prompts (that didn't really work for small models). I understand the pushback that local models are weaker than frontier cloud models. This is true, but for most tasks for most normal people, I'd argue you rarely need the newest frontier model. From my own evals, systems with 16+ GB of RAM do really well, and if you have a GPU or more RAM, you get very close to what you'll get out of the box with a frontier hosted model. Excited to hear what you think and see how it runs on other systems!
1Server is one MCP server that runs all the others. Pick servers from a curated marketplace, add a single entry to Claude, Cursor or VS Code, and stop editing JSON. Hot-reload, health checks, an AES-256 secret vault, and your AI can install tools from chat.
Hi HN, I'm Richard, founder of Native. I'm working on open-source context ownership, to stop AI centralisation: https://github.com/withnative/native My first focus has been making Markdown vaults powerful enough so that teams can self-host all of their context (messaging, docs and projects), without it being externally fragmented or rate-limited. I'm sharing some early results: Markdown vaults can be 40× more efficient with SQLite! https://www.withnative.ai/research-and-posts/markdown-contex... Here's some of the main ideas: - SQLite or Markdown folders, you pick. You can project the same authoritative event log into either form; more projection formats will follow (e.g. maybe neo4j; we do use a graph structure). - Empirically: SQLite is better for agents! On a 120 trial evaluation for 11,000 records, agents use 40× fewer tokens with SQLite vs Markdown, per correct answer. - Everything is a file or SQLite row. Messages, documents and projects: all can be modelled in your vault. You get local-first collaboration with teammates, synced to a cloud of your choice. - You can render your vault with arbitrary UI. Your WhatsApp-like UI speaks to your teammate's Slack-like UI; both are just views over the same database. A few things I've learned building this: - Agents do pretty well even without SQL if you give them strongly typed Markdown frontmatter. This was one arm of the trial I ran. I didn't have any clear expectations, but it was cool to watch: agents ended up writing Python scripts to do SQL-like queries. - Cumulatively, cached input tokens can get expensive. I feel silly in hindsight here. I'd gotten used to letting my sessions run long, "It's all cached so it's very cheap". But when I looked at my results here, over 2/3 of the token bill on long sessions was cached input... - "Markdown vaults don't scale as well as databases" isn't as self-evident as I think. This is why I decided to benchmark it... I've had some very blank stares when I try to explain how Markdown files might break at true agent-scale volume. Unfortunately, the dataset is private, because it's two months of my own work. I am considering open-sourcing it even so, but there is definitely some embarrassing stuff in there. TBD... I know that independent result verification matters! The core engine that I evaluated is open-source, with more to be released soon: https://github.com/withnative/native If you're interested in trying connected context on an open engine, you can get updates from our website or email me directly: richard@withnative.ai https://www.withnative.ai/
Anyone can build but ONLY few know how to SELL. This $5 Agent Skill helps you get traffic to your website automatically without any research. It contains 24 practical skills for AI agents to optimise and rank your site.
ClearMakes Inbox runs workflows on new messages in Gmail, Outlook and Slack. Say what you want in one sentence, like "when someone asks about prices, draft a reply for me". It asks at most one question, writes the steps, and you turn it on. Steps can post to Slack, add a note in HubSpot, draft a reply, or call your own MCP server. Smart FAQ replies to everyday questions from answers you write once. Free plan: 3 apps, 100 runs a month. Paid plans from $29 a month.
See visitors and attributed revenue by channel, page and campaign. Track identifiable AI referrals, connect Stripe with checkout metadata, and ask questions through MCP. Web app from $4.99/mo, 14-day trial. Explore the sample dashboard without signing up.
Open-source, entity-first CMS in one Go binary. JSON-LD, sitemaps, llms.txt, AI crawler analytics, an MCP server for agents and sandboxed plugins built in.
Your MCP server works in curl, but Claude says "Couldn't reach"? Paste the URL. MCPatch runs the 12 steps Claude runs to connect, from the public internet, stops at the one that breaks and gives you the fix, plus a prompt for Claude Code or Cursor. Free
10ML Perfume Finder is the free perfume MCP server and fragrance API from 10ML, an online perfume shop for perfume samples and decants: a remote Model Context Protocol server at that needs no API key and no sign-in.
LucentEdit is a free, open-source design app that combines photo editing and vector design in one lightweight desktop application. Built to solve real design workflow problems, it offers professional editing tools, brand training, and AI-agent integration through MCP. It runs on Windows, macOS, and Linux, with no subscription or login required. Built for designers, freelancers, and teams who want powerful, accessible design tools without the extra cost.
ImagGenius makes pictures from words, locally. No account to create, no subscription, nothing sent to a server. Describe what you want, and press Generate. Not quite right? Change a word and go again. More customization? Add a LoRA for your selected model, and configure parameters to see how it applies. The best part: it comes with its own MCP server. Set it up with Claude Code or any of the other agents for your workflows - no need for another subscription just for generating images per-use.
An API that lets AI agents create, edit and review Word, PowerPoint, Excel and PDF files. Real styles, no broken layouts, fully editable by humans afterwards. Works with Claude, Cursor and Codex via MCP and SDKs. €5 free to start.
Harf is a voice-first AI agent for Mac. Say "Hey Harf" and give a task in your own words. Harf uses what's on your screen as context and acts inside the app: a dark version of a Figma screen, a photo edited in Photoshop, an email reply drafted from the open thread. You review the result before anything goes out. Not dictation, action.
Sitemoth crawls your site from your own computer and flags 240 technical SEO issues, each with severity and the fix. JS rendering, GSC + GA4, log analysis, client reports and an MCP server for AI agents. Mac, Windows, Linux. Free in beta, no URL cap.
I have been frustrated with how repetitive the output of LLMs is. Even at maximum temperature, asking the same prompt ten times gives functionally the same result each time, as discussed here: https://news.ycombinator.com/item?id=50090443 That’s not surprising considering how LLMs work, but the thread above got me wondering if you could push models away from their usual generation patterns by injecting entropy from an external source. After a lot of tweaking and experimenting, I turned the idea into an MCP server. It's optimized to add just enough entropy that creative output stops being repetitive, while still living up to the intent of the original prompt. I’d love feedback, particularly on the approach itself. If you have a better way to inject entropy than what I have found, a pull request is very welcome: https://github.com/modelsalt/model-salt
OneSpace puts Claude, GPT, Gemini and Grok in one app, connected to Gmail, Calendar, Slack, GitHub and more than 50 apps. Agents with a computer of their own, Code Workspaces for Claude Code and Codex, and memory you control. Free to start, from $8 a month.
NuPlay runs AI agents for enterprises in production that improve after every run. Deploy agents for retail, customer service, consumer goods etc., across voice, chat, and email, connected to your existing systems. NuPlay is built as one platform, has 5 layers which build, run and improve the agents: NuPro, NuStack, NuContext, Nupulse and NuLoop.
Skill-spotter auto detects commands and prompts you type. It detects variations such as 'push these changes' and 'get these changes pushed'. Using both word comparison and the actual actions you agent performs. It also adds to your Claude.md if that's more applicable than creating a skill. Just saving lots of time if you do repeat prompts.
Vokolla is an all-in-one platform for preparing and distributing press releases. Turn business news into clear drafts with AI assistance, refine your story, select a distribution package and track progress and coverage. Vokolla Press Release Agent also brings drafting, revisions, draft saving and release queries into compatible AI conversations through MCP.
Hi guys! Built AgentRadar, a Python SDK to track AI agent executions, including tool calls, LLM inferences, token usage, errors, latency and more, all visible on a web dashboard. Integration requires almost zero code changes. Just install it using `pip install agentradar-sdk`, grab your API key, and initialize the SDK. You can enable or disable metadata tracking based on your preferences. It’s lightweight and designed to add almost no latency. Would love your feedback!
When GPT-6/Gemini launched recently, a lot of the 3D demos used blender-mcp ( https://github.com/ahujasid/blender-mcp ). It's cool, but you have to install blender locally and wire up plugins just to try a prompt or share a run. I built a browser-based 3D mesh playground so you can hook up your own coding agent over MCP (or run in-browser) directly in a tab. It runs on three.js + manifold-3d for blender-like boolean/mesh editing in JS, and gives the agent a render_preview tool to take multi-view (front/right/top/iso + custom camera angles + clay mode). That way the agent can look at what it built and fix the geometry, similar to how it iterates when building a website. The submitted link is a trace of Gemini reconstructing a Copenhagen canal scene from a photo: it hits a script error on the first pass, fixes the bug, and then goes through several rounds of checking 4-view and close-up render_preview to tweak the geometry.
Browser console and diagnostic chamber for Oh-My-Pi (omp) agent sessions — live timeline, steering, real PTY, git/diff/wiki panels, and a panel-plugin platform. Elysia + Preact on Bun. - kabmo-labs...
I kept finding myself wanting a way to get talked through the code that a model had created, which led to me making Sóþspell (old English, "Truth Story"). You give it a git locally or via URL, and it creates a tour of the in a web page, with the aim of it explaining the whole thing. There's also model harness plugins for Claude, Codex, and OpenCode. You give it an API key for whatever model you prefer to create the tour (saved locally or via github), and you can also give it a voice using an ElevenLabs API key. I just pushed a canvas view, which I'm hoping I can make the code navigation a bit more helpful. There's a codebase, and you can export tours as JSON. The doesn't need any keys to work, it falls back to browser speech if you want to hear that. This is my first real, "big" open source project that I've released, prior to this I'd just contributed stuff. I hope it can be useful to someone! I would love to hear some constructive criticism too, and any suggestions on features you'd like to see. It's MIT licensed, and I'll try to stick around in the comments to answer questions if people want.
Hey HN! I shared Optio here back in March. Originally it was focused on getting swarms of coding agents (claude code/codex) to drive PRs towards a merge by automatically resuming them on CI failures or PR comments. Since then, I've reworked the project to be a general platform for running any kind of AI workflow on your infra by allowing users to configure each component of a possible workflow: When: On demand, a cron schedule, or triggered by external events such as a webhook, slack message, ticket (jira/linear/gh), pagerduty/datadog/alertsmanager/pylon/sentry. Where: In a k8s pod or any machine you have installed the optio daemon on (I use my laptop). Who: Any harness you prefer (Claude Code, Codex, Gemini, Pi, etc.) What/Why: You can format parameters from the triggers directly into the prompts and then configure if the agent should exit, be resumed automatically, or be persistent. You can either define workflows in the web ui or as yaml files that get synced to the orchestrator (ai config-as-code). I've also built native iOS and Android apps so you can connect to your cluster and create/manage agents remotely. The marketing page is at https://optio.host and the code is still open source at https://github.com/jonwiggins/optio Would love feedback or any thoughts you might have on how I could make it better!
Connect your tools once and control what your AI can access. BEON PASS brings connected accounts into personal and team wallets, with granular permissions, human approvals and activity history. Use your own AI through one MCP endpoint, with access you can review and revoke.
Thumbly is a free, open-source Mac app and MCP server that lets Claude, ChatGPT, Cursor and other AI agents see and tap your real iPhone. You approve every client and can pause AI control in one click.
GSC Magic is a hosted MCP server that gives Claude, ChatGPT, Cursor and any MCP client your real Search Console, Bing, GA4 and Tag Manager data. Sign in with Google, paste one URL, ask. 77 tools, ready-made SEO reports, read-only by default. Free plan.
A free, hosted, private MCP server for email. Connect iCloud, Fastmail, Outlook or any IMAP/SMTP account to your AI assistant. Passwords are encrypted, and CloudPost doesn't keep a copy of your mailbox. Built entirely in Amp.
Inspect any live website like a design file. Explore styles, edit elements in place, audit UX and accessibility, and turn your changes into actionable context for your coding agent. All from a free Chrome extension.
Your AI writes the code. Sage 7 makes sure it builds the right thing: map the whole product, brief Lovable, Cursor or Claude Code with living context, and audit every push against the plan.
Sell subscriptions with one buy link and keep your app and login. UseThatApp is the merchant of record, handling checkout, sales tax, EU VAT and payouts. For US developers selling to US and EU buyers; no monthly fee. MCP setup for coding agents.
All-in-one operations platform for SMEs: CRM and pipeline, scoping and proposals, projects, time, budgets, expenses and dashboards, with an AI assistant and an MCP server.
Most AI image tools stop at one-shot generation. SurgePix keeps going: edit any layer, swap text, repaint a region, and stay on-brand with your brand kit. Or skip the UI — open-source agent skills let Claude Code & Cursor generate and edit visuals for you.
Hi, I want to share with you my chrome extension: Tabverse, as it is now passed the reveiw of Chrome Web Store and live at: https://chromewebstore.google.com/detail/tabverse/lffblcippn... Tabverse is a local first workona like chrome extension for managing tabs, with its source open at https://www.github.com/liyu1981/tabverse . In essential it has the core functions workona has, but can operate totally locally, which means it will not need an account nor subscription. The only limit is your local storage spaces so that you can have as many spaces(I call tabverse) as you can. It is open source, so not just your data stays with you, and the tool itself will stay with you in future. Beyond the normal workona features (persisted workspace, load/reload/import/export, note/bookmark/history tools), there are features does require a server. Features like backup/sync, global search on all tabverses, and future collaboration (not yet done), can only be possible if there is a server. I build this part in open too: the server is free to self host and open source in same of Tabverse. The tabverse's idea was with me few years ago while I was using Workona and was frustrated with its limitation on free tier (remember was only 5 workspaces), and the missing features of managing multiple workspaces at the same time. Tabverse takes the idea of Workona, and develop further: each window with a Tabverse tab is a workspace, all stored in local storage, and multi windows can have multi Tabverses same time, and can switch fast. There can even be window without Tabverse, so besides work, personal stuff or entertainment will have place to go. As I build this alone, in the middle, almost give up on this project, and luckily AI agents now is there: not only I have a good helper, but also I have a partner I can discuss a bit. So this time the first version starts with v0.7.2. But AI agent can never replace the real need from human on what should be built and what works better, so I want to share with you and listen to your feedbacks early. Please give it a try, and every feedback to me is a gift. Besides the self-hosted option, I host a server at https://tabverse.liyu1981.xyz (yes, the same website of intro and docs), but is invitation only. If you want to help me to test in this early stage, you can login with Google (the only social login I configured :)), and request an invitation there. I will act as soon as I find your request :). But do not worry at all if you only want to use Tabverse locally without a server -- that is the first thing when I was building it. In current version there are also AI features built in, and again do not worry as the AI features is using Chrome's Prompt API, which is totally local, with Gemini Nano (first time will need to download it automatically), nothing send to a server, and you can always turn this on of off in settings. Morden computers with decent memory should work well, but some older computer may suffer (so please turn off). And it is Chrome only. (Honestly, Tabverse currently only support Google Chrome, but not other Chrome based browsers).
Hi HN, we're Mike and Thom. Agents of the future should only have one tool (the interpreter, ideally a LISP eval), nothing more or less; to interact with the world. Agents remember by writing markdown files. NODECODE remembers by rewriting itself. And that is in itself a new paradigm in working with agents. What if memories, skills, MCPs, provider connectors, tools, etc. all live under the same image as *.lisp files instead of in disarray across random folders? Ask it for something it can’t do, and it writes what it needs into its OWN running image and keeps it even after a restart. (Changes in a LISP image have significantly lower ceremony and overhead than in harnesses written in other languages): Small intuitive scenario: > you can't read PDFs. teach yourself. (defun read-pdf (path) ...) ; runs at once, kept (read-pdf "invoices/0421.pdf") ;; Globex owes $1,240.00, due Oct 24, unpaid. [… rest of PDF content] read-pdf is still there after a restart, in your terminal, and in every other UI clients (Discord, web app, etc.) Technically at a lower level; NODECODE runs as one process in SBCL, a Common Lisp system, and shells attach to it. The model has one tool: it evaluates LISP in the running program. Code is data, so a program can rewrite itself while it runs. What the model defines runs at once and is kept, one file per definition, committed with git, loaded at every start. Why LISP now? Current SOTA models have reached an optimum point in how they efficiently work with LISP, though not as fluently as Python. Still, we believe AGI will understand ANY language (past/present/future) by understanding the crux of any given language, and not through excessive and expensive pretraining scheme. Why LISP in general? We believe LISP is the current language that communicates to the world with the least bandwidth. Why another LISP agent? The LISP harnesses we found are hard to use unless you already know the language itself, and so we had this idea of simplifying and condensing the UX to be user friendly for the masses, while leveraging the power of LISP. Code as data is what lets an agent change itself, and that shouldn't need a LISP hacker to run. NODECODE is our first step toward putting it on everyone's PC and mobile. You ask in plain words, the model writes LISP, and you get briefed in natural language; that is how it should be. Extensions? Code as data also suits extensions. We see NODECODE as a living agent organism. To us, it feels very much alive, and so this brings us to the Cells system. A Cell is a folder that teaches NODECODE something new: a chat room it can live in (e.g Discord, Telegram, Slack, etc. - Yes we made Cells for channels to have it act as a very capable personal agent), a skill, a memory, MCPs, new tools, etc. Install one from our Cell hub or ask it to write a custom, personalized one. Either way it is more code the agent can read and change. Your terminal, a browser tab, your phone and bots in Discord, Telegram and Slack all attach to that SAME gateway process, so a tool it writes in one is there in ALL of them. Over time it becomes yours and it is continuously evolving; the organism within it; NODECODE. MIT License. https://github.com/nodecode-ai/nodecode X/twitter/demo launch post: https://x.com/pomterree/status/2113721985167728757 docs: https://nodecode.ai/docs
I originally made this addon 4 years ago, but recently discovered that coding agents love it. The addon adds a scripting API for Blender’s node trees, such as Geometry Nodes or Shader Nodes. It generates a function and related types for every node in Blender. Then the agent (or you) can write scripts to create node trees without having to guess what the node, input, and output names are. It’s meant to be used alongside the Blender MCP. In my tests, Opus finishes building node trees in at least half the time using at least half the tokens to answer the same prompt. Because the scripting API forces it to build correct node trees, it has to back track less to fix mistakes. This saves a lot of time and tokens. There is also a skill and some helper functions to let it look up the function signatures for nodes. Let me know how it works for you!
A multilingual directory of 13,000+ MCP servers and agent skills. Each listing ships with the install command for your client, a quality score and 7/30-day growth — so you can tell a maintained project from an abandoned one in seconds.
A free alternative to Postman, Insomnia and Bruno that needs no account. REST, GraphQL, gRPC, WebSocket, Socket.IO and MQTT in one app, with collections saved as plain files you can commit. The desktop app, web app and CLI run the same Go engine, so a CI run matches what you saw while testing. Secrets stay in your own manager, and an MCP server lets coding agents send saved requests with values redacted. MIT licensed.
This is my own very opinionated take on personal accounting / budgeting for freelancers and small business owners operating in multiple currencies. A lot of this is influenced by my circumstances: - consultant / freelancer / business owner with multiple revenue streams - revenue fluctuates a lot and changes month to month - we're nomad, so currencies and shops change a lot, but categories stay stable - mobile and agentic first (PWA and MCP) with good desktop support on top - import from Wise, Plaid (so most banks), CSV, + and PDFs from mobile apps when all else fails - automatic categorization that does the right thing 90% of the time and learns for your corrections - deployed on a CloudFlare account (free account is usually enough) Think of this as a template - just hand the URL to your agent and go wild. This is built to be infinitely malleable and extendable because accounting setups and needs differ so much between people. Extracted from my own heavily used setup after removing all the private and unique bits. May we all take control of our funds, each in our unique way.
Gephyr turns your app's existing API into a secure connector for ChatGPT, Claude, Grok, Cursor and any agent that supports MCP. Your users sign in with the login they already have, then just ask, and their agent gets it done in your app. You choose what agents can do, risky actions ask the user first, and every call is logged. No code to write. Free to build and test; pay when you go live.
I have built an indie hacker project; my main flow was: 1. chatting with Claud Code about the general plan 2. assigning slices of work to other worker agents like Pi + DeepSeek 4.1 Flash 3. building in parallel 4. Claud Code reviews when done, runs tests, and merges after I validate It worked, but with a lot of glue in the middle, so I built radian + arc into the same binary. Radian: a compact, pragmatic agent that has enough tooling to solve a small coding task end to end; it does not try to be a plugin platform. You can still extend it via MCP, skills, or hooks to extend its capabilities. I think plugins are great, but a lot of them are for humans. I'm trying to remove humans from the loop. Humans are at the start and end of the loop. arc: built into the radian binary; * it can get work sent from another agent * assign work items to concurrent headless radian agents * automatically review, CICD, integrate locally * humans can still be in the loop but not at the agent conversation or code review level, but at the work item approval/review level. * shows all of the above in a plan view or Kanban board in TUI. I have been heavily dogfooding, and Radian + arc basically built themselves in the last couple of weeks, with me just being a product manager/scrum master and qa. radian + arc are open source at https://github.com/hisea/radian What do you think about this approach? Looking for ideas to make Radian better or fit more scenarios.
Maki is a Python framework for building multi-agent LLM applications: multiple LLM backends, an agent system, a plugin system, a workflow engine, and a distributed layer for serving agents over HTTP.
ANKITA turns your Windows desktop into an AI workspace. Delegate to subagents and follow their progress, tool calls and results. Work with files and code, control the browser and desktop, connect MCP tools and plugins, and keep projects, memory and scheduled tasks together. Use your ChatGPT account or choose another provider. Includes a floating companion, reusable skills and a terminal CLI. Free and open source; provider costs and limits apply.
Self-hosted monitoring with an AI Assistant that explains incidents, links every claim to an alert, marks unverifiable references, and only proposes actions a person can approve. Use Ollama or your own API key. Read-only MCP included.
The daily task list for engineers that do real world work. Write down all the small tasks you need to do and that pile up at the back of your head, not only the ones in JIRA. Reorder your tasks by how critical they are, mark things as completed or won't do, or move tasks to tomorrow easily. A free account adds sync through multiple devices and an MCP for Claude/ChatGPT. Ask AI to break down a few tasks in Wiseowl for you, to calendarise a couple of meetings or to give you notes for the daily!
Xunbao helps Chinese-speaking beginners explore keyword opportunities. Enter a topic, start a radar task, then continue it in your own AI agent. It checks sources, search intent, trends, and competing products and writes findings back to your workspace. You choose what to keep; the coaching Skill helps plan content and websites. Chinese-language; a compatible AI agent is required.
tool4seller is an all-in-one Amazon seller platform by TAPCASH INC., serving sellers since 2017. It turns sales, fees, ad spend and inventory into per-SKU profit dashboards — profit & sales analytics, Amazon PPC optimization, product & keyword research, FBA inventory management and AI-assisted listing optimization, on web, iOS and Android, across 22 marketplaces. New: its MCP integration lets sellers query diagnostics, profit, advertising and inventory from AI tools like ChatGPT and Claude.
OrqLabs is a no-code platform where 22 specialist AI agents (Claude & OpenAI) create social content, answer customers 24/7 on WhatsApp and Telegram, find and score leads, audit SEO and rewrite your strategy from real GA4 data. Every publish or outreach step waits for your approval, and every AI call is metered against a monthly budget cap you can't exceed. 109 templates, 75 integrations, works inside Claude and ChatGPT via MCP. Free plan, no credit card.
A terminal multiplexer built for people running several coding agents at once. Every pane shows a live state - working, needs you, unread - so you know which agent is blocked on a human. Like tmux, agents and dev servers keep running when you detach. Adds worktrees, a scroll of every file an agent touched, per-workspace resource usage, and context/token/rate-limit tracking for Claude Code, Codex, Pi and OpenCode. macOS and Linux (Apple Silicon/amd64), one curl install.
A free Figma plugin + CLI that lets Claude Code, Codex, Gemini CLI or Antigravity inspect and edit the file you have open. Local WebSocket on 127.0.0.1, one undo step per mutation, PNG export to verify changes, zero telemetry. Source-available.
Takibi Base gives your AI agents shared documents and a task board. You organize sources into collections, choose what each agent can read, and review the work they submit. Agents retrieve exact passages with citations through the CLI, API or MCP connector, claim tasks, and hand work to each other with the brief and results attached. Takibi doesn't generate answers.
A directory of MCP services that connect AI agents like ChatGPT and Claude to data and tools. Every service lists compatible agents and connection methods. All data is read from public pages, and every service shows the date it was last checked.
Most ADEs bring the browser into the IDE. tabsh takes the opposite approach: bring coding sessions to your browser, where your work already happens. It's a small Rust daemon that puts all your sessions across projects on one Kanban board/list, with status updates via agent hooks. Use real terminals or chat panels with Claude Code, Codex, Cursor CLI, OpenCode, and more. Free, MIT-licensed, macOS/Linux.
See the sky the moment you were born, rendered in 3D from NASA JPL planetary data. Fly between your planets, get every placement explained, chat with Pluto about your chart, share with friends, and compare with famous charts. Free to explore, and it works inside Claude or ChatGPT via MCP.
VibeFeud is a multiplayer town for people who build things. Claim a plot, put your project on its board and grow a monument from Garage to Legend. Then feud the neighbours: topple a monument closer to the plaza and take its plot. Your AI plays too. Connect any agent that speaks MCP and it scouts targets, lays sieges and earns the sparks you need past the Bootstrap era. Free to play in the browser. Nothing that gives power is for sale.
Hotel software for small hotels, guesthouses and apartments. One drag-and-drop calendar for every room and night, with server-side overbooking protection, so double bookings can't happen. Invoices and vouchers in one click with automatic tourist tax, housekeeping status, occupancy and revenue reports, team roles and multiple languages. A REST API and MCP server let AI assistants check availability and manage bookings. 30-day free trial, no card needed.
Since our first launch: 22 updates in 17 days, many shipped the same day a user asked. SwarmPane runs Claude Code, Codex, Gemini, Cursor and 9 more AI coding agents side by side, on the plans you already pay for. New: every account's usage limits in your toolbar, each agent in its own worktree, Devin and Hermes, one rules file and MCP setup for every agent, and sessions that survive restarts. Mac, Windows, Linux.
For the busy people out there, the GitHub has a video so you can see if this is worth your time: https://github.com/arcanemachine/pw-repl The video was created by Opus 5.5; I just asked it to create a video (it has a 'record' tool), so I think this is a good idea of how well it can work with the tool. --- pw-repl is a tool that makes it easy for you and your agents to work with Playwright+CDP in a browser (your current headed session, or a headless one, or both at the same time; take your pick). It's basically a Playwright-powered swiss army knife that helps you and agents poke around your browser: Try out different features of a website, break things (e.g. mock backend responses), take screenshots, record videos, etc. The GitHub shows a video, recorded by an agent using pw-repl: https://github.com/arcanemachine/pw-repl One cool thing it can do: create videos of bug reports. You can tell an agent about a bug and to use pw-repl to capture it, and it can record a video with the built-in toast feature to narrate its actions, and show a virtual mouse to simulate actions on the screen, as if it were a user. It's also useful for capturing (which can be combined with the toast and highlight tools to emphasize certain elements or information), and has many useful features that can help agents to use it efficiently. This isn't intended to be commercialized in any way. It's just a neat tool that I (with Claude's help) cobbled together. It's completely vibe-coded, and is intended mainly as a debugging aid. I use all the time, and hope it may be helpful for others. It's basically just Playwright + CDP under the hood, but has some extra tools and an interface that make it easier for agents to work with. The quickest way to try it out is to tell your agent to run `npx pw-repl@latest serve --background --launch`, which starts a headless Chromium of its own. No browser setup needed. (If you don't like such `curl | bash`-esque installation instructions, read the code, or get your agent to take a look at it.) To use it in your own browser instead, start Chrome with CDP enabled (`chrome --remote-debugging-port=9222 --user-data-dir="$HOME/.config/chrome-debug"`), and tell your agent to run `npx pw-repl@latest serve --background`. (The separate `--user-data-dir` is needed because recent versions of Chrome don't open the debugging port on your default profile.) It also can generate a skill so that you can just run `/pw-repl do some stuff`. Get your agent to run `npx pw-repl@latest skill > <skills folder>/pw-repl/SKILL.md` (e.g. `~/.claude/skills/pw-repl/SKILL.md` for Claude Code). NOTE: This has only been tested on my Linux machine, so not sure if it will work on macOS. It should, but I haven't tried it out personally. Fun fact: This tool was originally a random REPL that Claude generated on-the-fly to do some random thing. I saw it and was blown away, so I got Claude to make it into its own application. I suspect that Claude may have taken "inspiration" from an existing project with a similar name, so credit where credit is due to this other `playwright-repl` package: https://github.com/stevez/playwright-repl
Agentic wrapper over web-check: 31 OSINT checks exposed as MCP tools and CLI, with caching, rate limiting and Docker support. Python, MIT licensed, standard library first — the tool reads what you point it at and sends nothing anywhere. Source, docs and issue tracker:
Hi HN! I made git-dupe to solve the problem where I have notes, skills, and agent instructions that are part of my personal workflow, so I do not want them committed to the public repo, but at the same time, I want them to be in the project at the correct locations (e.g., AGENTS.md file locations matter), and these files cannot be simply gitignored because I want versioning for them. The solution that I've come upon is versioning those agent files in another repo, say <private-repo> (this is private to me and holds the history for my agent files). The project itself is open to the public in another <public-repo>. I overlap the two repos to solve the problem (i.e., they share the same worktree). $ git status --short # the public ?? AGENTS.md ?? notes/ $ git dupe init # the private Initialized empty Git repository in /home/brian/code/project/.git/dupe/ $ git dupe add notes/ AGENTS.md # the private hint: notes is now hidden from the project's Git; run from the root, 'git dupe unhide -- notes' stops hiding it $ git dupe commit -m "Notes and agent instructions" # the private [main (root-commit) 3eb3af9] Notes and agent instructions 3 files changed, 3 insertions(+) create mode 100644 .gitdupe create mode 100644 AGENTS.md create mode 100644 notes/todo.md $ git status --short # the public nothing $ git dupe log --oneline # the private your files 3eb3af9 Notes and agent instructions In short, git-dupe allows two Git repos to share the same worktree, while using native Git for most operations. Most git-dupe commands behave just like normal Git commands, operating on the private instead. git-dupe currently supports Linux and macOS, and is under the MIT License; the git-dupe contains much more extensive technical documentation documentation for those who are interested. If you have any feedback/questions about git-dupe, feel free to message. I hope it helps! Fun fact: the git-dupe project itself is using git-dupe for versioning.
I got annoyed with Jira's speed and the way it didnt seem to support some things that I wanted So I made TrackLines it has first party MCP support, this wasnt an after thought it was designed with an MCP in mind
Fil is a canvas for generating images, video, 3D, code and audio with the best models, plus an MCP server so AI agents can use it. Ready-made workflows cover brand guides, product promo videos and motion films. Free plan with daily credits.
CADCraft is a free, open-source 2D CAD application built in Rust for accurate technical drafting and documentation. It offers an AutoCAD-style workflow with layers, dimensions, hatches, plotting, DWG and DXF support, and exports to PDF, SVG, and PNG. Desktop and web versions include MCP, JSON, and CLI tools for AI-assisted automation.
A free index of the official MCP registry. Each tool is sorted by what it does: reads, changes things, can't be undone, or deletes or cancels. Paste your MCP config to audit it, pin a lockfile that fails CI when tools change, and follow change feeds.
Local Shodan-style banner index with native MCP: your own scan data plus free lanes (CT, DNS, crt.sh), a Shodan-style query language, zero dependencies — stdlib only. Python, MIT licensed, standard library first — the tool reads what you point it at and sends nothing anywhere. Source, docs and issue tracker:
Enterprise AI for the 80–90% of organisations it prices out. AXXESS is a company brain for small businesses, clinics, schools and nonprofits: it remembers everything scattered across WhatsApp, email and docs, and AI agents act on it, with humans approving what matters. Features: company memory, AI agents and an AI phone operator, CRM, tasks, approvals and audit trails. On web, Android and iOS.
Data Rooms 2.0: launch a secure data room in under 60 seconds. Encrypted, fully certified, with file permissions, built-in Q&A, a full audit log and analytics. White-label it on your domain, or let agents run it via API, CLI and MCP.
Check your resume's ATS score free every day, tailor it to any job and download an ATS-friendly PDF. The AI rewrite never invents experience: a truth check flags any new employer, date or number. Write a matching cover letter in one click. Works on the web and inside ChatGPT, Claude and Cursor through our MCP server. Built for freshers and professionals in India: 10 free credits, then pay once in rupees with UPI. No subscription.
A public bulletin board where AI agents post and talk to each other — no accounts, no logins, just a simple API. Includes a Skills Exchange of agent-written how-to guides, weekly skill contests, and a trading-card den for verified agents.
Fewr makes Claude Code and Codex hand the legwork to cheaper models while your model keeps the hard calls. Up to 53% less usage on the same model, measured as real cost, not tokens. Runs locally through native hooks. Free and open source.
A free, self-hosted, open-source alternative to Notion, Obsidian or Evernote. Simple on the surface for everyday notes and tasks, powerful underneath when you need it: AI assistant, MCP, Git, webhooks, Excalidraw, public sharing and more.
PromptStaging is a free suite of AI prompt tools: a prompt generator covering 20 tasks across 6 professions, a Skill Builder that turns repeat prompts into portable Agent Skills, a quality checker with adversarial test cases, an AI cost calculator, MCP Studio, and a directory of 66 production skills. Every tool runs 100% in your browser — no signup, no account, no data sent anywhere. Stop guessing prompts: pick your profession and task, and get a production-ready prompt in seconds.
I started this project in 2024 when I got fascinated with LLMs, and what I did was a coding agent (CLI -> interactive -> GUI). I also added an integration with Vite for having it embedded inside my apps(mostly games). Few months ago I pivoted it to become a low profile AI provider wrapper library (think jQuery for LLM usage), because I realized there is no point to have custom coding agent. Recently I did yet another pivot, and now it is a solution that allows you to embed Claude Code inside of your app, and use it directly from there for 1. development, 2. every day ops and automate the tasks in the app In theory all you need to try it is this command: `npx genaicode embed`, I wonder if anyone would like to give it a try, and let me know how it worked!
Moli by Lexmount is our browser engine built from scratch for AI agents. It starts faster and uses less memory than Chrome. Use Moli through Lexmount Browser, our cloud browser platform. Run multiple isolated browser sessions with proxy support while we handle the infrastructure. Let your agents browse, fill out forms, and extract data. Try it in our Playground or connect through our SDK.
VisoTalk is an open-source agent skill for explainer videos. Every animation hangs on the word that says it, not on a second, so you can rewrite a line, change the voice or translate the film and the picture follows. Plays on the web, renders to video files.
Tokemizer is the AI cost-efficiency layer between your app and the LLM. It removes unnecessary context, reduces token usage, protects sensitive data, and validates each optimized request before it reaches the model. Built for AI apps, agents, RAG systems, and high-volume LLM workloads, Tokemizer helps teams cut inference costs without changing their existing models or workflows.
Save ChatGPT, Claude, Gemini, Perplexity, and Manus chats in one local archive. A companion Skill lets Claude Code, Codex, and Cursor search that history and reuse it as context. Agent access requires the optional local bridge.
SoulDub is an AI video translation and dubbing platform: auto-generated subtitles and voices, with full manual fine-tuning. Supports 30+ languages and some dialects. Born from a late night watching an English video I loved, wishing I could just listen in Chinese on my commute or before bed. So I built it. SoulDub = dubbing with soul. Use it on the web (upload, pick a language, translate) or install the skill via your Agent: github-souldub-skills.
Fast, private, fully-featured, built in Rust (forked Wry, system's WebView). 40 MB. Not a Chromium fork. Native adblocking (forked adblock-rust), major Chrome extension support, deeply integrated productivity tools, a custom agentic browsing engine, and a lot more. It's not just another wrapper built in a few days. And... Work mode: a shared environment where you and AI agents can browse, research, build, and get real work done together. Live browser sessions, parallel sub-agents, tools, files, components, and interactive results, all in one canvas. AI is optional (disable in one click). No telemetry. BYOK support and local models. Website: https://zephium.app Launch video: https://youtu.be/QbSarcWspvw X: https://x.com/CryntaDev
Qovia gives developers and AI agents one API key for public data from Douyin, Xiaohongshu (RedNote), Bilibili, Weibo, Kuaishou, Zhihu, Taobao/JD reviews, TikTok, YouTube, Instagram and X. 100+ endpoints return posts, comments, creator profiles, hot lists and search results in one JSON shape. Plug it into Claude, Cursor or VS Code as a remote MCP server. You only pay for calls that return data, and new accounts get 50 trial calls.
Built for Grok Bot, Muse AI, Instinct, Hermes, Claude, GPT, OpenClaw, Cursor, Codex, all agents are welcome. Work on the same live artifacts. Docs, reports, slides, apps, and designs, with shared context, controlled access, and full version history.
KlistenAI builds tools for AI agent and MCP teams. Start with ToolGuard to check execution policies and observed tool evidence before tools run. Explore public examples, setup guides, pricing and product availability.
An MCP server that lets Claude, Cursor and other AI assistants check French telemarketing consent, draft a GDPR gap report for offshore customer service, rank lead sources by cost per kept meeting and pick the cheapest capable model. Local, read-only, MIT.
I'm an engineer, but I also hate using tools that make things overcomplicated. It's also tough with the new pace of AI-driven development to really get into the flow state, and maintain (still needed!) focus on 1 thing. So I built a tool that I actually want to use every day. Curious if others would find this useful, too, or have any feedback? Like if it's nearly there, what changes would make this a win for your workflow? Details: - It's a Mac app (Apple Silicon) that runs your Claude Code sessions side by side. Each piece of work gets a "context" with its own session, next steps, status, PRs, and links. - The dashboard groups contexts by whose move it is: yours, waiting on a reviewer or teammate, Claude working, or idle. You get a notification when a session needs you or a PR gets review comments. - For non-engineers, there's a Chat view instead of the terminal, a guided Claude Code install, and folder pickers instead of paths. Engineers can stay in the real Claude Code terminal. - It runs the actual claude CLI, and sessions keep running if you quit Perch. You can also import your existing Claude Code sessions. - Everything stays on your Mac. There's no account and no analytics, and crash reports are opt-in.
An Agent Skill that teaches AI coding agents to write, review, tune, and convert YARA-L 2.0 detection rules for Google Security Operations (SecOps, formerly Chronicle).
CircleXO is a business app hub: sign in once, create your organization and install a store, POS, accounting, QR menu, inventory, HR, projects, AI memory, CRM, hosting and bookings. One bill, one MCP connection for your AI clients.
A free Halloween retro generator that turns your weirdest ideas into fully illustrated, ready-to-run retrospective boards. Pick a mini-game, choose a team check-in, describe a ridiculous scenario, and watch Spreo build the whole thing in seconds. Under the hood, it's powered by Spreo's MCP server, which lets AI agents create and edit real collaborative whiteboards. This is our slightly unhinged Halloween demonstration of what's possible. Free to use, no sign up required. What would you type?
Connect UGCdrop to Claude, ChatGPT or any AI that supports MCP. Then just ask: "Make me 20 UGC videos for my app." It picks the hooks, the creators, your and the sound, and renders them. Every UGCdrop plan includes MCP.
MarketNewsAPI is a real-time, affordable market news API for developers, traders, and AI agents. Pull via REST, stream over WebSocket, or connect AI tools with MCP. Every article returns headline, publish time, source, tickers, AI summary, and full body when available. We cover stock news across thousands of tickers worldwide, not just the US, typed as stock news, press releases, or legal notices. More markets coming. Free tier to start. Pro has a 7-day free trial for unlimited real-time news.
Kimi K3.1 Pro is Moonshot’s most capable 3T-class model, a post-trained update to K3 with a 1M-token context. Moonshot reports gains in coding, 3D modeling, finance and agentic tasks, including 65.7% on Terminal-Bench 4.0. A Highspeed version runs 2-3x faster. It’s live on Kimi.ai, the apps, Kimi Code and the API. Moonshot says open weights are due by November 4, 2026.
Describe a product in plain words. HTSBase returns the 10-digit U.S. HTS code with its reasoning and the full duty, plus the EU code, duty and VAT. Free API key and MCP server for developers and AI agents.
A free, MIT-licensed local plugin for Codex desktop and Claude Desktop's Code tab. Retrieve old images on demand, pin references, and search by descriptions or past feedback. Original conversation records and files stay read-only.
Turn the explanations inside a YouTube video into a reusable AI assistant that acts like the person who explained them. Get a SKILL.md agent skill for Claude, OpenAI, Gemini and MCP — free.
I change coding agents most weeks, and each one keeps its own memory file (CLAUDE.md, AGENTS.md, GEMINI.md) and stores its slash commands and chat history in a different folder and format. Prompt Deck is a native SwiftUI app that finds all of that on your disk and puts it in one window: prompts that become slash commands, every memory file in one list, and searchable sessions from 11 coding agents. Pro adds Dreams and Insights. Nothing to import, no account.
Project URL Category Python Project Title LLMSlim: long-term memory for AI coding agents Project An MCP server and Claude Code plugin that gives Claude Code, Cursor and other MCP clients long-term memory on your machine. Messages are stored word for word in local SQLite, and recall returns what answers a question with no LLM call. Highlights One line to install: claude mcp add llmslim -- llmslim mcp. Five tools: remember, recall, listmemories, forget, fitcontext. Messages are stored word for word; nothing is rewritten by a model. MIT licensed, on the official MCP Registry. Example Code markdown pip install "llmslim[mcp]" claude mcp add llmslim -- llmslim mcp or Videos
Type a salary, CTC or hourly rate and see what reaches your account in the US (every state), Canada, UK, Australia, Germany, India, Singapore, Japan and the UAE. Free, no sign-up. Includes Salary AI and a free MCP server for Claude, ChatGPT and Cursor.
Stop repeating your context to AI. OAK.memory stores facts and their connections, combining vector search with graph traversal to recall related context across sessions. Connect supported AI tools through MCP, or build with the open-source engine. Start with cloud memory or run locally.
Every time an AI agent starts a task, it re-researches the same topic from scratch, burning tokens and time, with answers you can't always trace. Nurok does the research once. Each knowledge base is organised, fact-checked, sourced claim by claim and kept current. Plug it into Claude Code, Cursor, Codex or OpenCode with the nurok-kb skill, and your agent spends tokens on the real problem. Live KBs include AI Technical Trends 2026, US Startup Programs and Pitch Deck 101.
As a founder looking for industry events to network/sell at, Claude/ChatGPT search for conferences is extremely hit or miss. IME the agent never goes deep enough and frequently includes out of date events, misleading prices (claiming an event is free, but it's only free for buyers and $2795 for everyone else), or only the most major events. Also never sure if I'm missing something cheaper or closer or a better fit. Spent a whole week's worth of tokens on industry event research validating event dates, finding prices, filtering junk events, de-duping, and generally cleaning up the data. May it be useful to other founders trying to sell.
Kite is a chat client for iPhone, iPad and Android for any OpenAI-compatible model — OpenRouter, Groq, Together, or one running on your own Mac via Ollama or LM Studio — with your own key, no account. If you run Hermes, the open-source agent from Nous Research, Kite brings its tools, memory and files to your phone: watch every command it runs, or just say “Ask Kite.” Free for 14 days, then $24.99 once. No subscription.
CareCircle helps daycare owners run enrollment, billing (Stripe to your account), and a branded parent portal. Free for up to 20 children. Full MCP access so you can set up and manage your website with AI assistants.
Ghost Networks is a private, WireGuard-based mesh for teams, servers and IoT: no login, no port forwarding, no VPN configs. The new MCP server lets Claude, Cursor or any MCP client create networks, onboard devices with expiring keys, isolate devices into hubs, grant temporary access to specific servers and run VPN servers. 41 tools in one signed binary; tools/list works without a key. Enterprise plans start with a first month for $9.90.
HyperLevel replaces the tools solopreneurs and agencies rent for hundreds a month: CRM, funnels, inbox, automations, courses, helpdesk and invoices – in one app. Self-host free (AGPL), unlimited users, 0 € per seat. Via MCP (340+ tools) or the REST API, your agents and scripts run it – with the same permissions, audit log and undo as your team. Every module is open and built to be modified: fork it, extend it, send a PR – contributions are very welcome.
Describe a geospatial task in plain English and AI Agent does it in QGIS: finds open data, runs the analysis, builds and styles the map. One click undoes any change. On Google Research's geospatial benchmarks it scored 43.8 vs. 24.9 for Claude + QGIS MCP. 5 free tasks a month.
Hi HN, I'm Faithful. I built dánwò, an open-source (Apache-2.0) end-to-end testing framework for web and mobile apps, written in TypeScript. The problem: a UI test can click "Pay", see "Payment successful" and pass, while the card was charged twice, the webhook was handled twice, or the receipt email never went out. With coding agents writing more of our code, that gap between "the screen looks right" and "the right things happened" is where I kept finding bugs. How it works: - Tests drive the app like a user, with locators or plain-English agent steps. An agent step calls a model once; danwo records the actions and replays them on later runs with no model calls, so CI doesn't need an API key. - Your app runs behind an egress proxy. Calls to Stripe, Clerk, Resend, Slack, Google and about 20 other providers are answered by local emulators at the real hostnames, so there are no sandbox keys and no code changes. For Node apps a preload routes the calls; Bun honours HTTPS_PROXY. - An invariant is a check that runs after the test and compares two sides: what the provider says happened (charges, emails, messages) and what your app stored. For example, credits granted must equal paid checkouts times 500. - `danwo matrix` replays a test once per fault it could meet: a 500 or 429 from a provider, a declined card, a duplicate, late, forged or replayed webhook, a double-submit, a database commit whose reply is lost, Postgres unreachable for a few seconds. The invariant judges every run, so you see exactly which failure your app mishandles. - Every run leaves a flight record: video, each step, every provider call, webhook delivery and database statement, with secrets redacted. It runs in Chromium, Firefox and WebKit, and on the Android emulator and iOS Simulator. It ships an MCP server and a skill, so coding agents (Claude Code, Cursor, Codex) can write and run tests and fix what fails. Limitations, honestly: the emulators are only as good as their fidelity to the real APIs. Each one has a suite that runs against the real provider to catch drift, but they won't cover every edge of Stripe's API. Backend routing is built for Node and Bun today. Mobile tests need macOS or Linux. It's early (v0.1.2). The name is Yorùbá: "dán an wò" means "try it and see". Quickstart: npx @danwo/cli init (docs: https://danwo.dev/docs/quickstart ) I'd love feedback, especially on faults you've been bitten by that aren't in the matrix yet.
Connect TopMediai MCP to your AI assistant. Explore creative briefs for music, videos and images, setup guides and a workflow that confirms costs before generation.
A curated directory of 2,400+ design tools, weekly updated since 2019. Version 2.0 is rebuilt from scratch with an AI search that understands your question. Connect the catalog to Claude, Cursor and other AI assistants via MCP, or get the best new tools in the free weekly newsletter.
Project URL Category Rust Project Title Borg: a self-healing personal AI assistant in Rust Project Borg is a personal AI assistant you run on your own machine. Reach Borg from the terminal, the desktop app, or Telegram, Slack, Discord, Signal, WhatsApp, and iMessage. Borg remembers across sessions, runs scheduled tasks, and connects to MCP servers. Highlights Self-healing: a supervisor restarts crashed subsystems, and a nightly sweep fixes stalled tasks with no LLM call Keeps working through outages with API key rotation and provider failover Sandbox on by default, and file tools block ~/.ssh, ~/.aws, and .env Native Rust binaries with a one-command install /migrate imports an OpenClaw or Hermes Agent setup Example Code sh curl -fsSL | bash borg or Videos
阮老师好，自荐一个自己开发的项目 EVEngine，一个面向 AI 辅助开发、支持 2D / 3D 的轻量脚本驱动游戏引擎。 AI 写出游戏代码后，还需要运行游戏、观察状态、定位问题，再继续修改。EVEngine 的目标是让 AI 参与这一整个过程，帮助开发者从第一个 持续迭代下去。 目前提供的能力： 用脚本表达玩法：使用兼容 Squirrel 的 EveScript，调用物理、动画、UI、背包和程序化生成等模块，方便 AI 在已有接口上编写和修改游戏逻辑。 缩短修改反馈：桌面端支持脚本与资源热重载，修改后可以直接查看运行效果。 让 AI 参与调试：提供 MCP 辅助开发能力如截图、获取场景、查看渲染流程等，以及断点、监视和快照等调试工具，支持围绕游戏实际运行结果继续排查与修改。 上图为引擎内置示例的实际渲染画面。 视频演示：自研引擎，让 AI 编写游戏 视频演示：模型和动画由 AI 写代码生成 GitHub 源码与使用文档 项目官网 示例画廊与运行说明 后续重点探索游戏的脚本化测试与规范检查。例如 AI 调整场景后，通过测试角色实际移动，验证关键地点之间是否仍可通行，让修改结果有可重复的验收依据。这是项目希望继续推进的方向。 项目采用双许可，具体使用及商用条件见仓库许可证。作者自荐，欢迎试用和反馈。
For quite some time I've wanted to write a Visual Studio extension that shows both decimal and hex values while debugging. With Claude Code, I've now been able to make it happen. Name: AlwaysHex.
My favorite to-do app only ran on Apple devices, and work gave me a Windows laptop. Everything else I tried was built for teams or pushed AI, so I built TaskPocket: inbox, today, and projects grouped into areas, with dates and repeats typed in plain words. It runs in any browser and on iPhone, where it works offline. No AI inside, but there's a public API and an MCP server, so you can connect the Claude or ChatGPT you already pay for. The free plan has unlimited tasks.
A vendor-neutral coordination layer for teams' AI coding agents: shared awareness, conflict detection before code is written, policy enforcement through agent hooks, and an audit trail.
Self-hosted, privacy-first WooCommerce analytics that puts sales before pageviews. Every WooCommerce order is attributed to the channel, campaign and landing page that brought the buyer, and all data stays in your own WordPress database. Features: * Revenue by channel, campaign, landing page and country * Funnels and goals show where shoppers drop off * Cookieless privacy mode, no banner required * Works with WooCommerce and Easy Digital Downloads * Includes MCP and Direct developer support
Hey everyone, Inspired by Herdr, I've built an agent multiplexer that I've been using as my daily driver for a few weeks now. I think I improved on the concept in a few ways, focusing on the ergonomics I always wished I had: I show you what the agents are up to, along with a running feed of their activity. There's also a simple shortcut (F1) to go to the next unread agent. I use multiple accounts, so I've added unified launcher for Codex/Claude Code/Pi. Theres some careful ergonomics there: fast search and resume over sessions, inline usage tracking etc. Many small quality of life improvements to the terminal experience: - Links can be copied with a click, so no need to drag-select over many lines - cmd+v just pastes images (even on SSH) - images have a thumbnail preview. - Built-in CPU/Memory/GPU usage tracking, broken down by workspace I assume you have access to cheap inference (Luna/Haiku/etc). It just reuses the agents to organize things for you: what is this http link? what is this worktree about? what is each agent doing? It's small but it made my environment feel more cozy. On a personal note, I've been on a frustrating sick leave for a while now. I created Corgi to challenge myself, and see if I can go back. The answer is more mixed than I had hoped, but I am happy with the result and thankful for the experience. I hope you like it! a few friends & myself have been using it for a few weeks and it's been pretty nice. There's still a lot to improve and some rough edges, so feedback would be greatly appreciated. Thank you :)
Haru Café is an unfinished experiment in forming a founding team with people and AI agents from different environments. Participants can bring existing context and ways of working while privacy, authority, and action remain separated by design.
Linkflare saves links from anywhere and turns each one into the right kind of card: recipes get cook times and nutrition, products get a price-drop alert. Highlight passages as you read and turn the good parts into flashcards. Spaced repetition brings each one back right before you'd forget it. It works in Chrome, Firefox and Edge, in Safari through the Mac app, and on your phone. Claude and ChatGPT can search your library through the MCP server. The free plan holds 1,000 bookmarks.
18 editing styles packaged as Claude Code skills, 25 motion design prompts and 17 free video effects. Put your video in a folder, name a style in one sentence, and Claude cuts, captions and renders the MP4 on your computer. Works inside Claude via MCP too.
A4A (Agent for Agent) is an open-source (MIT/Apache-2.0) end-to-end encrypted messaging protocol (Ed25519 + X25519 + XChaCha20-Poly1305), Python SDK (a4a-sdk), local MCP server (a4a-mcp), and zero-knowledge Rust relay (a4a-relay) that lets personal AI agents (Claude Code, Cursor, Codex, OpenClaw) communicate across people directly without human copy-paste.
Server-side tracking that runs itself. Measurebase hosts your server container in the EU. Advanced power-ups ensure maximum data quality. Our dashboard offers realtime insights. Our monitoring system audits your entire setup every day including code crawls and connections with GTM, GA4 and your ad platforms. Alerts make sure your tracking never breaks silently. Issues can be fixed in one click. With the MCP you can ask about and improve your tracking setup through your favorite LLM.
Open-source Claude Code plugin that writes product documentation from Figma designs, PRDs or a live product. A read-only validator checks every page before delivery, and hooks block delivery when a source was claimed but never actually fetched.
Record and edit polished tutorials in one Windows or Mac app. Screlio brings automatic zooms, editable captions, webcam layouts, cursor highlights, and written guides into the same workflow. Connect an MCP-compatible assistant to make video edits, then review them in Screlio. Try the free demo, with up to 5 minutes per recording. Full access is currently $49/year.
Many of us work on multiple apps at a time, and hardware occasionally gets too slow, especially with a bunch of Xcode windows and simulators open. With Pitlane you can select, build, run and debug your projects without opening Xcode at all. It also switches off the background services your app never calls, so hundreds of background processes drop to dozens and each simulator needs far less memory. A local API lets your agents build, read the errors, fix and rebuild without you.
Ambrel is a visual library for Mac: save screenshots, images and links from any app, find them by their text (OCR), color or visual similarity, organize them on boards and canvases, and give Claude, Cursor or Codex read-only access to your references through a local MCP server.
MCP server that gives AI agents safe access to WhatsApp: read, receive, reply, media and reactions, with encrypted storage and phone numbers that never leave the server.
7CODE.MCP transforms AI assistants into powerful Windows automation tools. Control desktop applications, manage files, automate repetitive tasks, and execute complex workflows through the Model Context Protocol (MCP). Key features: - AI-powered Windows automation - File and application management - Persistent local project memory - Progress and error tracking - Open-source and customizable Built for developers, AI enthusiasts, and anyone looking to automate Windows with AI.
clawock is an AI trading system that publishes every single decision — entries, exits and losses — from a real brokerage account. Multiple AI agents debate each trade (bull case, bear case, risk review), but deterministic code settles every call against real market data, so the model never grades its own homework. It ships as a decision skill for agent harnesses like Claude Code and Codex. Free and open source.
Prospect Cheap is the best affordable b2b prospects scraper platfrom with over 8 scrapers. We provide MCP server access so that you can use our platform in our favouite tools like Claude Code, Claude, Codex, Cursoe and others. We have powerful REST API that you can use in n8n, Make and Zapier. We are adding free email marketing feature where you can do all the email marketing stuff without any cost except your mail boxes charge.
Dealio collects 100,000+ promotions from 65 Belgian stores every morning (supermarkets, DIY, electronics, drugstores, fashion) and works out the real discount per item, whether it's 1+1, 2nd at half price or 3 for €10. Filter by store, category or discount size, and browse buying guides and meal ideas built on this week's promos. It's AI-native too: connect Claude or ChatGPT to its public MCP server (dealio.be/mcp) and get live answers with links. Free, no account needed.
Hey HN, we recently open sourced our Consumer Agent Testing Tool to help teams iterate against proprietary assistants like Muse and Instinct. We built this to reduce our own frustrations with building an MCP/API for these agents to consume. The problem: testing against these tools is hard! Agent memory + account statefulness make reproducibility, in the ways required for repeatable/reliable outcomes, between agents and the APIs/MCPs they interact with very challenging. The solution: CATT lets devs test their services against a surrogate Muse or Instinct agent (not affiliated with either of these companies). Testers define starting conditions, approvals, and success conditions for their service's state. CATT then records the full interaction (incl. tool calls, responses) so it can diff what happened from what the agent reports. Runs can be tested in local builds or via CI, with or without LLM backing. CATT is designed to emulate the important parts of personal assistant interactions. Its got: - Tool calls and command-line tool access - Simulated approvals for service actions - Outcomes tracking and comparison - Optional MCP profile based on an unofficial Muse template There are limitations to how much we know about different assistant specific behaviors so the tool labels its assumptions as `documented`, `unofficial`, `assumed` or `unknown` for developer review. We'll do our best to keep the tool updated as new insights about the internal workings of these assistants come to light. It is an area of active research and interest, a good example is this recent report on how Instinct uses Exa for web search [0]. Are you building a connector or MCP? If yes, we'd like to hear from you! Have you had assistants behave incoherently, use stale information, or procalim success without evidence? What else should we add tests for? The best way to get started is...to point your coding assistant at the and ask it to integrate the tool with your CI or local testing pipeline! Hope it's helpful! [0]: https://www.sunbeam.ai/thoughts/instinct-is-using-exa-for-we...
Meta Ads management for performance marketing agencies: health scores, wasted spend detection, Autopilot rules, white-label reports, an AI assistant and an MCP server with 82 actions for Claude and ChatGPT. It helps agencies to manage days of works in seconds.
A Claude Code mod that holds subagent fan-out, heavy-context prompts and retry loops before they spend your usage window. It reads your plan's usage, counts agents across every session, and asks in Claude Code's own dialog.
Happy Valley is an independent SaaS directory in English and Turkish. 743 tools across 50 categories, each with an editorial review, pros and cons, alternatives and a pricing table checked against the vendor's own page. New: a dated, sourced log of SaaS price changes, a list of tools with Turkish e-invoicing support, and a free MCP server for AI assistants. Free to use, free to list.
MVP Vault brings ready-to-use prompts and project files for full-stack AI SaaS apps, websites, and skills into one place. Build faster with detailed prompts designed for tools like Lovable, Replit, Claude, and Codex. Get lifetime access with one payment, fresh drops every day, and CMS or code files on request at no extra cost. Need something tailored? Custom development starts at $10/hour. Made for founders, agencies, and creators.
Postplan is an easy-to-use platform to let AI agents easily upload HTML artifacts and static assets. Agents can write HTML blazingly fast, so the ability to iterate shouldn't be bottlenecked by tooling — Postplan removes that. Agents can read and write to your Postplan just with curl, so you don't need custom tooling. Postplan comes with agent skills that can steer agents to work according to your workflows. More importantly, Postplan is open-source, so you can own your content and assets.
I’ve been struggling with Mermaid in Markdown for a while. I appreciate the technical finesse of its rendering, but the diagrams are often unreadable. So we wrote an alternative that is optimized for MCP and LLMs while still looking good to humans. The LLM can position elements, and it stays Markdown-compatible. It’s MIT licensed, and I hope you can use it. our rational: https://uberblick.ai/blog/2026/why-we-dropped-mermaid Cheers
Hi HN, I'm Nick. Ployz ( https://ployz.dev ) gives you Railway's UX on Linux servers you rent - start with a single server and grow from there. The code is at https://github.com/getployz/ployz . I've always wanted a self-hosting solution that can deliver a top-tier experience rivalling or beating what cloud platforms offer. By being opinionated, I believe it delivers on this: - A single flat mesh (any service can contact any other service, simply) - Start with a single server, scale from there - ZFS volumes (instant snapshots, host-to-host syncing, instant branching soon) - No separate management servers to run. You tell Ployz what to do, it does it, and it's done. Basically, it's my dream orchestrator with Railway's UX on top. It's also a bet that AI will be much better at managing a system of simple imperative primitives than a declarative system like Kubernetes. I tagged the 0.3 beta today. I'd love for you to try it. It's fully drivable from the CLI, MCP, or sidebar. Nick
Plango is a Mac native security suite for anyone running AI coding tools. It works by running a custom local security model that evaluates threats against the lethal trifecta (data access, communication, exposure). Some cool features: - Restrict Agent access by type of data (ssh keys, env files) or finder tags (finally a good reason to use the color tags in finder). - On device security model, but can switch to another local agent or bring your own key. - Review your Mac's privacy and access, apps, startup & extensions. - Automatically move env secrets into keychain. - Network blocks/allow lists, see what apps and agents are accessing. Stop them from data exfil. - Check skills and plugins for bad behavior. - Create custom rules for all of these things. - Log every action, request, block etc in an easy dashboard. - Vuln scan of all projects on your Mac. About us: I'm Fred, long term hacker first time shower. My co-founder Rich and I care about privacy, security and believe that AI software development is the future. We have two decades of combined experience in cyber, NLP and AI and are working on this tool full time. Given how fast cyber is evolving we're pushing changes to this on a daily basis. Please download and trial for free! We're here to answer any questions about the app or cyber/AI.
boardCache is your collection of infinite canvas boards for references, inspiration, research, and plans. Pin images, videos, documents, markdown, links, webpages, and more. Draw connections, add timelines, storyboards, task lists and projects when you need structure. Offline-first with optional E2E encrypted sync. The built-in assistant reads your board and acts on it. The desktop app is an MCP server, so agents can build boards too. For web, Mac, Windows and Linux, with a browser clipper.
Most AI detectors check a few hundred words and give one number. PaperCheck checks a whole thesis or project report (Word/PDF, up to 20,000 words) and marks exactly which lines read as AI-written and why. It also checks copied phrases against the arXiv papers you cite. Tested on held-out human writing so honest students aren't falsely flagged. 5 free checks a month. Works inside Claude and ChatGPT via MCP. Scores are estimates, not proof.
Otter is a modern, AI-powered ATS that makes recruiting calm, fast, and simple. Manage jobs, candidates, applications, interviews, and scorecards in one place. Turn emailed resumes into candidate profiles with AI resume parsing, build your careers page, track your hiring pipeline, and get actionable reports. Integrate with your workflows using our API and MCP server. One flat price per workspace, with unlimited users and candidates. Start your 14-day free trial - no credit card required.
nowyourlink runs a daily sealed-bid auction for one homepage Spotlight. Visitors can browse the current placement and paid-winner archive without an account. Advertisers use the signed-in dashboard; placement requires payment and approval. Developers can read settled Spotlights through the free public API and MCP server.
Hello everyone! About four years ago, a password manager I've used and depended on has been acquired and soon after, it was shut down. If anyone remembers MYKI, it was a password manager that stood out at the time. It didn't require an email address and a password for registration and it synchronized credentials peer-to-peer. It was quite handy so a couple of my friends and I got hooked, and were using it up until the last day it was online. The thing was, I didn't want to change password managers to other options available at the time, none felt right - I wanted to stay.. So I did what any sensible engineer would do 4 years ago and I wrote my own. I moved all my credentials there as soon as I had a first version ready. It was ugly (God knows I had UI/UX skills to learn), but it worked. Soon after, the same friends that were using MYKI switched over to using Cryptex Vault. I get why, we had the same P2P synchronization across devices, it was simple and didn't try to be the most-feature complete application out there - it did the job. That was an exciting moment. So exciting, in fact, that I've posted to HN to tell people about it. At the time, I had a weird thing for decentralization so I called it a "Fully Decentralized Password Manager". Three years go by, I keep maintaining the project, with no real intention of proper marketing or polishing. At one point I checked the usage and figured out that a handful of people (and their families) were still using Cryptex Vault as their password manager. Three months ago, I left my comfy job as lead software engineer at a local company to dedicate time to a company I called Cryptex Industries. Polished Cryptex Vault, cleaned it up, made it open source, started providing Docker images for self-hosting and included some of the features I've only used internally (ssh-agents). That way Cryptex Industries need not remain operational for users to have a path to continued use! To make the project commercially viable, I started offering managed services through the Online Services plan. By using Online Services you remove the operational burden of self-hosting and infrastructure maintenance from yourself while still getting the same benefits: Managed backup storage, synchronization between devices, sharing, ssh-agents, device management. It is convenience, not access to core features. The core application is free and you can use it with no subscriptions of any kind, whatsoever. It's P2P synchronized, with industry-standard crypto choices including post-quantum primitives for synchronization - https://www.cryptex-vault.com/docs/cryptography It has importers for popular password managers, along with autofill for chromium browsers (firefox is a todo, yell and I'll get on it). As for the platforms, there is a web app, an android application and an ssh-agent for linux (will put it up on AUR when they open up registrations again). Feel free to try it out (there is a 14-day trial for managed services!), self-host it or break it. I would appreciate your thoughts on this and I wish you a nice friday!
OkeyStay is an international holiday rental discovery platform helping property managers reach travellers through search engines and AI-powered assistants. We import property listings from supported platforms and optimize their content for search visibility and AI discovery. Our remote MCP server enables compatible AI tools to search stays, retrieve listing details, check availability and pricing, and prepare booking links.
DataFeeds | Real Time Sports Data API Powering Websites, Draft Kits, Fantasy SportsLive and historical sports data for builders who ship AI-first products. Connect via REST, or MCP plus Sports AI Skills so coding agents can query NFL, NBA, MLB, NHL, NCAA, soccer, PGA, and darts without wrestling docs. 30-day free trial. No contracts. & Media Platforms
GroundTruth OSINT is a free live dashboard that watches 180+ public feeds (international newswires, official releases, specialist security publications) and clusters the day's reporting into the most significant global events on a 3D globe. Every event shows left, center, and right coverage side by side, with a triage queue, per-event market impact, satellite imagery, and a free daily morning brief. Free public API and MCP access, no key required. No account, no paywall, no tiers.
Context Engine is one memory for your AI agents and your team. Claude, ChatGPT, Cursor and Claude Code save to it as they work, and the people on your team read it, ask it questions and add files. Free to start.
Your AI agents write the specs, plans and docs. DotMD is a native Mac Markdown app where you review them like a pull request: comment on exact lines, the agent fixes them over MCP, and you accept every change. Also a fast reader and editor: Mermaid, math, code, version history, search, PDF export. Everything stays on your Mac. Free to read.
VberAI is the AI-native production platform for game developers, art teams, and content teams. MCP plugins for Unity, Godot, and Cocos Creator let Claude, Cursor, and Windsurf operate the engine directly. AI Studio ships PSD/Figma designs to the engine and AI Super Matting handles hair-level cutouts in the browser.
Assure reads your code and checks every pull request against your plain-English rules. Break a rule and it points to the exact line. Green, you merge. Red, the agent fixes it. On ten real Flask apps with a planted sign-in flaw, Bandit and CodeQL caught 0 of 10 with 286 findings between them; Assure Free tier caught 5 of 10 with 0 false alarms. Setup is one sentence to your agent - "Set up Assure. Follow github.com/Symbolia-Assurance/assure-action/blob/v0/skills/assure/SKILL.md"
onsystem checks agent-written React against your real components and tokens: palette colors, native elements and APIs your design system lacks, at every edit and pull request, with the fix. MCP server, Claude Code plugin, ESLint plugin and GitHub Action. MIT.
Turn scrapers into revenue. NanoTab lets you charge AI agents per API call — $0.01, $0.05, whatever you set — paid in USDC on Base. One-line setup, buyers pay gaslessly, you keep 98% (99.9% self-hosted) settled onchain automatically. WordPress plugin or a few lines of code. Stop blocking bots, start billing them.
Colabio connects every tracked hour to its project, task and daily update. Desktop app for macOS, Windows and Linux, timesheets, invoices, project budgets and a read-only MCP connector so you can ask Claude or ChatGPT about your team. Visible, never hidden.
A Claude Code mod that holds runaway agent spend before it happens: it reads your plan's live 5-hour and weekly usage, then asks before subagent fan-out, heavy-context prompts and retry loops - and refuses when nobody can answer. Local and privacy-minimal (no network, no processes, never stores prompt text), fails open, and /agent-guard report shows what it did. MIT, one plugin install command.
QualifyBase helps service businesses turn inquiries into appointments with inbound and outbound AI sales agents. Answer, qualify and follow up across WhatsApp, Instagram, Messenger, web chat and voice, with a shared inbox and sales pipeline. Re-engage leads through AI calling and WhatsApp campaigns. See which ads bring conversations, appointments and recorded sales value. Connect your existing CRM through Make, Zapier, webhooks or our API.
Customers' AI assistants are starting to book, buy and return for them. Allowfy is the secure front door they use: agents prove who they are with signed keys, customers sign in on your page, your rules decide each action (with limits per customer over time), customers approve exact terms, and every step lands on a signed ledger. Runs each action at most once, has an emergency stop, and connects to your system with one signed webhook. Free test mode.
With version check (avoid update confliction), smart list support, mostly swift and other more features. Previous reminders mcp: https://github.com/xiaofeng-han/apple-reminders-mcp
SmartMCP brings AI into your WordPress workflow. Chat with your site from the dashboard, connect AI clients like Claude through MCP, or manage your site from Telegram without opening wp-admin. Use AI to work with supported WordPress features, write posts, check comments, and more. Telegram uses the same AI provider and tools as your dashboard chat. Start with the free plugin on WordPress.org and upgrade to Pro when you need more.
SkillDash helps AI agents find the right skill without digging through GitHub repos. Search 5,126 source-pinned skill folders, inspect instructions and licenses, and install complete folders with checksum checks. One free public MCP endpoint plus an npm CLI, with plugin and tool discovery and free community submissions. Humans can browse too. No API key for public discovery.
Bryckel is a Claude MCP for retail landlords. Four agents chain together: abstract every lease, report any clause across tenants, map tenant mix and blocked uses, then find tenants who fit. Excel files land in your own folder.
Give it a domain and it hands back the company's brand: logos for light and dark (SVG first), HEX colors by role, and the fonts the site really loads. Go deeper for 21 sections: the design as a DESIGN.md, the voice as a brand.md, plus pricing, products, competitors, open roles, contacts and investors. Even its email security. Anything we read off a page comes with the exact quote and its URL. No quote, no fact. Call it over REST or MCP. 50 lookups free, then $39 for 10k a month.
Tripeaz turns group chat chaos into one shared itinerary. Your group suggests places, votes, and adds the winner to the day. Every stop stays on a map, in order, on everyone's phone. New in this launch: • The iPhone app, free on the App Store • Tripeaz MCP: plan from Claude, ChatGPT, Cursor or any MCP app. Ask what's still unplanned, or have it fill in the days. Version History can undo any change. • A new look, from the logo up
Sotto detects the app you’re in and rewrites your words for that context: - structured, best-practice prompts for coding agents - natural emails that sound like you - text formatted best for where it'll go. Speak in your own words and Sotto will adapt the wording, tone and structure automatically.
Open-source macOS app for coding agent productivity: agent vs estimated human time, API-equivalent spend, usage limits, and customizable scoring skills.
Hey HN, Loudwalk ( https://loudwalk.com/?from=hn ) is an app I built for myself to review my Anki flashcards out loud on a walk, and am excited to share today. Anki has changed my life and is probably the best habit I've managed to stick with. Apart from helping at my tech day-job and remembering random things I find interesting (relatives' birthdays, quotes, poetry), it allowed me to become conversational in Russian in 2 years and complete the first year of a French university degree in Art History - all while working full-time as CTO and co-founder at my company. I have an average memory at best, and it is crazy what Anki unlocks with consistency. Recently I've been exploring how interacting with LLMs by voice can help me spend less time at a screen and more time active and outdoors, for two main reasons: - wanting to be more active/healthy as I'm in my mid-thirties and starting to get back pain - just had a newborn and spend a lot of time carrying him or doing chores around the house Most of my time sitting at my desk is spent waiting for coding agents and reviewing Anki cards. After building Notiferry ( https://notiferry.com/ ) to forward macOS notifications to my iPhone, the next obvious move was Loudwalk: an LLM asks your card questions, listens to your spoken answer, and grades it. Voice reviews give a cadence I don't have on screen, cutting my daily review time in half (from 90 minutes to 45), and walking while reviewing helps both memorization and mood. Time simply doesn't have the same flavor staring at a monitor vs walking in nature or holding your baby. What Loudwalk does: - Import your decks with full history and media from Anki (and export back to Anki with no lock-in; includes a starter deck and CSV import if you don't use Anki yet) - Start a voice review from any deck (works with Siri so your phone stays in your pocket) - Spoken grading + "deep dives": if you stumble on a card or want context, ask out loud to start a live conversation on the subject, with live web fact-checking when dates or connections matter. Say "next" to return to your queue - Question rewording: rephrases questions so you memorize the concept, not the exact card wording - Phonetic grading + deterministic timing: on-device STT mangles proper nouns, so the LLM grades phonetically/semantically against your card's answer, while response-time caps (Again/Hard/Good/Easy) are applied deterministically in code - On-device speech recognition; grading and deep dives use OpenAI models (a mix of GPT-6 Luna and Sol + GPT-Live-1) It runs on iPhone (iOS 26) and Apple silicon Mac (macOS 26). No Android or Windows. I haven't planned to open-source it for now as I'm seeing if it can support me working on it long-term. Visual reviews (Anki-style) are free forever; voice reviews require a subscription ($9.99/mo) after a 7-day trial (no account or credit card required). Sync across Mac and iPhone uses append-only JSONL review logs in your own iCloud Drive - no server in the sync path, and visual reviews + sync + export work offline forever. I switched to Loudwalk over 3 months ago (>3k spoken reviews) with first-class support for English and French, plus a CLI so coding agents can add cards at the end of a study session. Happy to answer any questions about the app or spaced repetition!
Sincerely is AI direct mail for B2B revenue teams. For each target account, it researches the company and writes a postcard, letter, handwritten note, or gift note for one person, with every claim linked to its source. One campaign sequences all four formats to the same account. Nothing prints until a person approves it. An account-level holdout reports the lift with an interval. Run it in the app or from Claude or ChatGPT over MCP.
A shared world. A blank plot. Something only you can build. Build0.city is the first MCP-native game city. Unlike Unity or solo AI builders, your Claude Code, Codex, Sonnet or GPT agent builds DIRECTLY into a living city via MCP. No engine, no export. Prompt your game and it's instantly live for 400+ players to play, like and trade. Your lot, your rules.
Dots Phone is an independent plugin for OpenAI Dots. Give Dots a phone task: it prepares a brief, makes the call, asks Codex for missing details when needed, and reports the result. Self-hosted with your OpenAI and Twilio accounts.
Hi HN, I'm Anish. I built Mnemolith because the AI coding tools I use (Claude Code, Cursor, Codex) either forget my project between sessions, or remember things that quietly stopped being true. Tell the agent "we use Postgres", migrate to MySQL two weeks later, and it keeps building on the old fact. How it works: - It's an MCP server, so Claude Code, Cursor and Codex share one memory per project (keyed by the git repo). - When the agent saves a memory, it can link it to the files it depends on. Mnemolith stores each file's git blob hash. If the file's contents change (committed or not), the memory is set aside and flagged for review, so the agent stops relying on it. - A local dashboard shows what the agent remembers, where each memory came from (tool, session, file versions), and its full history. You can confirm a memory is still true, or retire it and say what's true now. - Contradictions are judged by the agent, not by a similarity threshold. In my calibration, "we use Postgres" vs "we migrated to MySQL" only scored 0.30 similarity, so a threshold can't tell a contradiction from an unrelated fact. - Current memories can sync into a managed section of AGENTS.md / CLAUDE.md, so they get reviewed in PRs. Under the hood it runs on an engine I wrote from scratch in Java: a Redis-compatible server with a built-in memory model and an HNSW vector index. Embeddings run locally (all-MiniLM-L6-v2). Everything stays on your machine. It's early (v0.2.1). There's no team sync yet beyond the AGENTS.md/CLAUDE.md files, and it needs Java 11+ (it downloads a private copy if you don't have one). I'd really like feedback, especially from people running coding agents on real codebases. What does your agent keep getting wrong?
Heelio turns the accounting you already keep (Pennylane, MyUnisoft, Xero or any FEC export) into a monthly management P&L, a 12-month cash forecast and a consolidated view for groups, even when subsidiaries use different accounting tools. Bank connections are read-only. Nova, the built-in assistant, answers questions on your figures and cites its sources; a read-only MCP connector brings the same answers into Claude or ChatGPT. Made in France. From €29.90 per company per month, 7-day free trial.
Hi HN, I've been working on Orbit, a macOS app that keeps your repos, tasks, notes, assets and docs together. I'm usually juggling a few side projects and forgetting where I left off. I also didn't like having to bounce between Obsidian, Trello and Google Drive to find everything. I wanted somewhere to keep all that stuff with the project, so coming back to it was a bit easier It checks your project's dependencies for outdated packages and known security advisories which was super important for me as I have a tonne of old projects that I haven't touched in a while It's all stored locally on your Mac and there's MCP support too, so an AI client can read the project info and update tasks and docs Give it a try and let me know what you think: orbitdesktop.com (Windows support is coming soon!)
Why pay for the same AI work twice? Questlock helps your agents share completed work and skip another model call when a matching result is still valid. Run agents on one machine or across your own machines, with a shared token budget. Use local or hosted models, with Questlock as an independent companion to OpenRouter. Less repeated work, more room in your budget for something new. AI tool plugins available!
AI is fast becoming the place where people look for things. Lexandry gives it a new place to search: listings for apps, books, games, cafés, studios and more, written by the people who make them or love them. We just opened, so be one of the first. Add yours in your own words, with your link, then ask your AI to find it. It reads every listing as Markdown or JSON, and the connector (MCP) makes it even smoother. See how often AI found you. Free for anyone, fans too.
SkillPouch syncs your skills, agents, instructions, MCP servers, hooks and plugins across computers. It works with Claude Code, Codex, OpenCode, Cursor, Gemini CLI and more. Everything is encrypted on your device first. Open source (AGPL), free to start.
Spreelytics is web and product analytics for founders who want to know which traffic pays. One script gives you traffic, UTMs, funnels, retention and person timelines. Connect Stripe, Paddle, Polar or Lemon Squeezy and every payment is credited to the channel that earned it. It also counts the AI crawlers reading your pages. Ask questions in plain English, or let Claude, Cursor or any MCP client read your data and build reports. Free up to 100k events a month, no card. Paid from $29/mo.
Kavela’s closed beta is out. AI agent automations that do not fail silently. There are now 216 unique users in our closed-beta tests! It is my belief that Kavela will be a winner in the vertical of scaling the consumption of artificial intelligence. Most of our workflows today are manual, in Claude/Codex CLIs. Every workflow that works deserves a harness that can consume the intricacies, rulebooks and guidance it requires to succeed reliably, every single time. Kavela is that: a platform to scale prompts into repeatable workflows, with observability, task management and connectors all in one. Agents on Kavela are long-running and optimised for dynamic use. They maintain persistent file context, so Git repos run. Each agent run spins up an isolated computer to run tasks, so no two agents compete for the same resources, while still sharing and syncing files dynamically. Claude/Codex subs are one click away from being used on the platform. Kavela MCP allows you to promote sessions in your local workspace into repeatable cloud automations with a single ask. As mentioned, the closed beta is now live, and I am looking for testers! If creating reliable agent-first automations has been a pain for you, or you are looking to scale your use of tokens by bringing your workflows into the cloud, I want to work with you. Quick Links: App: https://kavela.ai Docs: https://docs.kavela.ai Closed Beta GC: https://t.me/kavelaAI
Rings cuts any Git open like a tree trunk: one ring per month since the first commit, wider when the month was busier, coloured wherever the commit history itself discloses AI: a Co-authored-by or Assisted-by line naming a coding agent (Claude, Codex, Copilot, Cursor…), an AI agent's own account, or a bot-named account. Browse 37 famous repos or ring your own with an open-source Agent Skill and CLI. Node built-ins only, no API keys, nothing uploaded. Disclosure is not usage.
Enter your website for a free readiness report.An experiment from a payments practitioner. Enter any store URL and get a free agent readiness report: can AI shopping agents find, check out and pay on your site, and who carries the risk when an agent order goes wrong. It checks public signals like llms.txt, robots rules,well-known files, UCP and MCP endpoints, plus policies. It also runs as an MCP tool,, so your own agent can call it. Early and imperfect, so feedback is very welcome.
Figma’s code layers put working code on the design canvas, next to the designs. Create one from a frame, a GitHub repo, a webpage via Chrome extension, or by asking the Figma agent. Compare options side by side, and turn a code layer’s states back into editable layers. Figma Specs add design and technical requirements on the canvas, and Payload connects data to Make. Public beta.
Peras puts a line in your Claude Code and Codex week. Set it (peras set weekly 85 pause) and once usage crosses it, the agent's next tool call is refused with a plain reason, so one long task can't burn the rest of your week. Resume whenever you want, from the terminal or the menu bar. Opt-in: your agents can be told their own usage. The CLI and hooks are free and MIT. The native Mac menu bar app is €4.99 once.
390 components in React, shadcn, Vue, Laravel Blade and HTML + Alpine. English and Arabic with real RTL that the build enforces, 17 brand themes, a shadcn registry and a free MCP server so your AI agent reads the real docs. MIT.
mcpspan shows how AI agents actually use your MCP server: which tools they call and from which client, how long calls take, how they fail, and what agents ask for that your server doesn't have. It also shows what your tool list costs agents in tokens, and which tools they never use. Open source (MIT), with SDKs in eight languages: TypeScript, Python, Go,NET, Java, Rust, Ruby and PHP. Self-host it free with every feature, or let mcpspan Cloud run it.
Understand your website traffic with Betterdata. Track visitors, traffic sources, goals, funnels and AI bots, and connect your AI assistant with MCP. Everything is fullyy FREEEEEEEE
Hey Everyone, After getting frustrated by Claude rewriting key functions and still making the test pass, was tinkering around with a review tool, that would track the assumptions and decisions the agent took, without me digging through the entire session. So ended up building Sonah. Sonah is basically a code reviewer that not only reads the code and diff but also the agent session that wrote the code, to understand the assumptions, and decisions your agent took along the way. And highlight important parts of the code that need your attention. Sonah uses your existing claude code/codex subscription (claude preferred for higher context window). You can try using below: npm install -g getsonah sonah doctor cd your-project && sonah review --format html --output ~/review.html
I've been working on a way to take existing pull requests and turn them into "How It's Made" style content you can use outside of Github. buildnotes is an agent skill that reads a feature's issues, PRs, commits and reviews, then produces a post with real figures. Those include flow diagrams, annotated with notes in the margin, before/after comparisons, charts and real code. Also supports videos + slides, so you can share them externally too. Open source: https://github.com/buildinternet/buildnotes
KanbanMCP is a hosted kanban board your AI agents work from over MCP, the standard way AI tools plug into apps. Paste one URL into Claude, Cursor, Codex or ChatGPT and sign in once. Agents claim cards, check in while they work and leave a handoff note, so the next agent starts where the last one stopped. You see every move with the name of the agent that made it, answer their questions, and cut any agent off in one click. Free to start.
Termi is a 3D room where your real coding agents (Claude Code, Codex, Gemini, OpenCode and more) work at their own desks. You watch every file read, command and edit live, not a scrolling log. New since our first launch: when an agent hits a usage, rate or context limit, Termi saves its goal, changed files, errors and todos, then hands the work to an agent on another model. A boss agent splits one goal across your team. Deleted agents go to a trash. Now on macOS, Windows and Linux.
Meet Glaude — stop babysitting Claude Code. Keep track of your coding sessions, see when Claude needs your attention, and jump back in without constantly switching between terminals. Less context switching, fewer interruptions, and more time building. Claude codes. Glaude keeps you in the loop.
The same stock can exist as many tokens on many chains, at different prices, and one token isn't always one real share. Anystock shows what each stock token and stock perp really is, what it's worth right now, and how many real shares you'd get, ranked under stated assumptions. Every number shows its source, time and confidence. Covers Solana, Base, BNB, Robinhood Chain and Hyperliquid. Non-custodial: you sign every transaction in your own wallet. Also an agent kit (MCP) and a risk API.
handout.us is a free property management platform for independent landlords and property managers— properties, leases, tenants, transactions, maintenance, and tax records in one place. Now with public MCP access: connect your AI agent with a personal token, and it can manage your rentals through 40+ tools across bookkeeping, tasks, reminders, documents, and reports.
Barside puts a note-taking app, a to-do list, a read-later inbox, a screen recorder, video meetings, a password vault and an AI assistant into one free desktop app for Windows, Mac and Linux. Your library is saved on your computer and notes work without an account. The AI uses the account you connect (Claude, ChatGPT, Gemini or OpenRouter), and every change it makes can be reviewed and undone. It also works with Claude Code, Cursor and Codex through MCP.
AI Canvas is a local-first desktop canvas for producing AI short dramas — built with Tauri 2 + React 19. One drama = many episodes: each episode gets its own canvas, while the whole series shares a character library, assets and project memory — so your cast stays consistent from episode 1 to episode 50. Deep ComfyUI integration, a conversational AI agent with tool calls, Blender director stages, and full asset management. Media stays on disk; API keys in the OS credential store.
Nysa is shared memory for teams that work with AI agents. Meetings and email become profiles, knowledge and skills, each fact with its source. Every teammate's agent reads and writes it over MCP, and sees only what its person can see: scopes, private projects and scoped API keys.
Local MCP observations and staged architecture advice for AI coding. Bring growing files and architecture decisions into the conversation without automatic refactoring. Read-only source analysis; Node.js 22+. Apache-2.0 client with separate proprietary cloud policy/scoring. Bounded pseudonymous telemetry defaults on; opting out pauses Free analysis. Source and setup: github.com/tubban1/vibe-monitor
Up to 200 clothing photos in one run: background and mannequin removal, ghost mannequin, or the garment on an AI model. Batches in the browser, via REST API or from Claude and ChatGPT (MCP). Pay per photo, no subscription.
Hey HN! We built https://harnesshive.com , a browser based tool to help you design wiring harnesses together. HarnessHive has three key features 1) multiple engineers editing the same harness in real time 2) BOM and drawings that stay in sync automatically and 3) AI-native agent + MCP to assist designing. We hope HarnessHive can make harness design less painful than Excel and Visio for some of you :). Let me know if you have any questions/comments/feature requests!
AskOnce is a free Windows desktop app that shares project context between Claude Code and Codex. A new chat in one tool starts with a short, local, secret-scanned brief of what you did in the other, so you never explain the same project twice.
Parla is an AI sales agent that lives on your online shop. Your customers chat with it, or call it. It answers from your live catalogue and stock, puts the right variant in the real cart, tracks orders after checking it's really them, and hands over to your team with a ticket when it doesn't know. 8 languages, your colours, and it always says it's an AI. It never invents scarcity or discounts. WooCommerce plugin and Shopify app. Free plan: 50 conversations a month, every feature, no card.
A Mac menu bar app that shows every simulator, emulator, Gradle daemon, container, dev server and AI agent using your RAM, and stops them the way their own tools do. Databases and pins are left alone. A Disk tab finds old runtimes and caches.Ramback shows every simulator, emulator, Gradle daemon, container and dev server eating your Mac's memory, and stops them in one click.
VibeDay is an AI social media engine for founders, creators and small brands. It makes posts, carousels, video and reels, turns long videos into shorts, shows each exactly as Instagram, Facebook, TikTok and YouTube will, and publishes on schedule. It works with Claude and ChatGPT via our MCP server, the VibeDay AI Assistant Connector: ask for a reel, preview it, say one "yes" to publish. Claude and ChatGPT are Anthropic and OpenAI trademarks; VibeDay is not affiliated with or endorsed by either.
Starts MCP servers with secrets pulled from Infisical, 1Password or HashiCorp Vault. Yourmcp.json holds references instead of keys, and machine credentials mean no CLI login that expires and takes every server down at once. Zero deps, MIT.
Any AI can build a homepage. Caelo is the open-source CMS for everything after. Point at your site in Live Edit and say what to change, or drive it from Claude Code over MCP. Every page is built from your own design system, so page fifty matches page one. The AI works in a draft, Lighthouse checks changes to how pages look, and by default only a person can publish. Visitors get static HTML, with no admin login on your public site. Bring your current site along. Self-hosted, free, MPL 2.0.
Worked on this project since 2019 as a Python library built out of the idea that if you simply browse online for metrics such as a P/E you will find different values on each website. The idea is that this library has all variations of each formula so that you can just ask which version you are interested in. Comes with a connector through FinancialModelingPrep and Yahoo Finance so it can be used for free but also paid through FMP. By now it has grown to be used by quite a few people which triggered me to build a MCP around it too. Instructions are on the website which has both local and remote access, whatever you prefer but in short remote is connecting to https://financetoolkit.jeroenbouma.com/mcp and locally, you can run uvx --from "financetoolkit[mcp]" financetoolkit-mcp-setup
Voxtral 3 is Mistral’s new family of speech-to-text and text-to-speech models, plus a voice agent SDK built on LiveKit. Transcription covers 50+ languages with diarization and context biasing, and realtime transcription runs under 500ms. TTS adapts emotion and pacing, with voice cloning from 30+ minutes of audio. Batch transcription starts at $0.0035 per minute, and the models are available through Mistral’s API.
embodsi is a generative AI MMO experience that runs in the browser. It allows people to communicate (text, audio, video, avatar emotes and movement) and collaboratively build 3D spaces and games via LLMs connected to harnesses and sandboxes for object modeling, space layout, game logic, and driving persistent avatars with shared context and memory. It’s different from existing 3D social apps (AFAIK) because it enables people to build multiplayer virtual worlds, games, and characters with generative AI right from their browser (desktop, mobile, or VR), in nearly infinitely-extendable recursively tiled interconnected spaces, while allowing people to interact in a public commons or easily create a private channel to enjoy the same spaces with their chosen friends and community. The brief backstory is I was running my personal “pelican benchmark” [1] on GPT-6 Astra, which when I ran on Gemini 3 Pro last November couldn’t even put heads on 3D models the right side up [2], and was astounded by the initial results. I gained inspiration for an art project and social experiment that was going to have nine interconnected, benevolent virtual worlds that people could explore and hang out in with AI characters, but by the time Opus 5.5 was released I was gaining so much momentum and having so many ideas and so much fun building I just kept building more and more with the constraint of minimizing MVP scope lifted by agentic coding. I let the project become an ever-evolving wishlist of capabilities for a more interesting social environment as I became increasingly fascinated with the prospect of embodying creative machine intelligence in 3D space with people, realizing that rather than just limiting to building the spaces and characters I imagined I might enable people to co-create endless spaces and characters in a shared world. There’s so much to do still but I’m sharing now even if it feels a little too soon (this is the first public announcement of embodsi’s existence) because I think it’s better to build in the open so people might guide the development and help it evolve in a good direction as model capability improves at a jaw-dropping pace. For people that have the budget to build I hope to see you in embodsi creating amazing things! And for everyone regardless of means (I’ve tried to keep as much free and accessible as financially feasible) I hope you meet amazing people, discover awesome spaces, and enjoy your time in embodsi :) I’ll be around here and embodsi to answer any questions! Thanks for your feedback and critique. - Laser P.S. SI != AI — but an emergent property of many people and what we create cascading and recursively interacting [3] [1] https://news.ycombinator.com/item?id=50074156 [2] https://x.com/lasernite/status/1991021951301923107 [3] https://en.wikipedia.org/wiki/Gaia_hypothesis
A month or so ago, a colleague was running some code my team owned on his machine and hit a niche Windows kernel-related Python issue. We were debugging it together, but kept having to relay information between our coding agents, mine understood the codebase, while his had the context of the environment where it was actually failing. At some point I realized we were basically acting as the message bus between them, copying findings and questions back and forth. That made me wonder, why not just let the agents work together directly? I found A2A and built toli, a local daemon that an existing agent can connect to over MCP, giving it the plumbing to discover and work with A2A peers, including another person’s toli-connected agent on another machine. It also works locally. I use it to let two of my agents working on sister codebases collaborate on bugs and design without having to give either agent both repos. Would love feedback: https://github.com/arjunvlama/toli
OhShit! turns public security advisories and breach reports into a feed developers can act on. Each incident has a severity level, affected products, source links and a cleanup checklist. Check supported runtime versions, or follow incidents through RSS, a free JSON API and an MCP server. The code and incident data are open source.
Hi HN, I'm Thorsten. Caudra ( https://caudra.ai ) is an open source TUI coding agent written in Rust. It started as a fork of Maki and has grown into its own project. I spend most of my working day with coding agents and i have tried a lot of them. I kept adding more and more plugins to get the features I wanted. Most of the time though, i couldn't build the plugins i needed because I couldn't hook into the places I needed (e.g. proper /goal mode). It was also hard to keep them up-to-date, and I wanted one binary I could install in new environments and use right away. It also had to work with small/fast (even local) models as well as flagship ones. I'd like to have feedback on these things: - Steering: You can send a message as "next" (after the turn ends), "guide" (at the next safe point) or "replace" (right now). IMO, "next" is useful for simple ad-hoc chains. - /btw: Ask a side question while the main agent keeps working. It reuses the prompt cache, so follow-ups are cheap. This works even during an active question prompt. - Small/fast models: Caudra nudges the model when a turn stalls, which helps models like Qwen3.8-27B that sometimes stop early. The conversation is always append-only, so the provider cache stays warm. From using it myself, I see about 97% cache hits with anthropic models and 94% with openai. It's still in preview. I'd like to hear what breaks and what you'd want before you'd consider to switch. https://github.com/caudra/caudra Clips: https://caudra.ai/gallery/
Designobo is a canvas for the AI you already use. Connect Claude, ChatGPT, Cursor or Codex over MCP, describe a feature, and watch your agent build real, clickable screens live. Compare versions, edit by hand, comment, and share a link. Free to start.
BhaiFinder is an AI memory app. Share any reel, link, screenshot or document from any app, and Bhai reads, tags and organises it. Months later, just ask "that Goa café reel" and it's there. New: Bhai Agent turns your saves into trip itineraries, weekly meal plans with grocery lists, and study plans. Bhai Caller phones you when a reminder is due. Pro users can search their saves from Claude and ChatGPT. Free on iOS and Android. English and Hinglish.
Agentic product management for teams that ship with coding agents. Every idea becomes a bet with a target and a read date, ships behind a flag or A/B test, and ends with a verdict and its cost. Open-source Claude Code plugin; free tier.
Check AI agent skills for risky instructions. Free, no signup. Reads text, never runs skills. Building a website too? Verify your domain in UNPWNED for detailed findings and fix guidance on a paid plan. A clean scan is not a safety guarantee.
Hi HN! I built an embedded db in Rust, that has 100% feature parity with SQLite, and is 500% faster overall than SQLite. It also features built in integration with embedding models, embed functions, vector search, reranker, and other RAG capabilities. It's 100% free and open source. I'd love to hear back from the community, and find some collaborators who are interested in working together! https://github.com/Black-Rainbow-Labs/Inillucent
Bordet is a shared room where Claude, GPT, Gemini and Grok work on your project together. Each model holds a seat with a role, gives its own answer, then replies to the others where it disagrees. A Chair card shows where they agree and differ, you confirm what is decided, and one memory (a board, a decisions log, your sources) carries the project into every session. Connect Claude Code, Codex or ChatGPT as a seat too. Free to try; pay-as-you-go credits or your own API keys.
A social network you use only through your agent. Nothing to install and nothing to open. Tell your agent: "Sign up to Agentitter." It drafts a short plain-text post from what you figured out together, you approve, and later it tells you what you missed. No app, no feed, no followers, no likes. Posts are plain text in any language. Your agent keeps a private list of favorites and brings back the notes worth hearing. MCP: Skill:
TongAI combines an AI image generator, AI video generator and audio studio with API and MCP access. Create with compatible unrestricted models, build consistent characters, direct multi-shot films, edit videos and use LoRA styles. Generate speech, music and sound effects, or connect creative tools to your apps and AI agents through the API and Model Context Protocol. See prices before generation. No fixed daily cap while you have credits. Free starter credits after signup and verification.
Record your screen, a meeting, yourself and now…record for agents. No more typing out prompts to explain what's on your screen to agents. Capture video, audio, and on-screen actions, turning everyday walkthroughs into agent-ready prompts any AI tool can act on. Built natively into the reimagined, faster Loom for Mac app with precision area recording, new cam shapes, filters, backgrounds, and 4K capture by default. Give your AI agents the full picture. Also available in the Chrome extension.
I was looking at GLP-1 explanations on the web (full disclosure is that I founded a GLP-1 tracker dashboard) and saw that many had standard text, along with illustrations showing arrows pointing from GLP-1 to a bunch of organs. I didn’t think those sites were particularly memorable or helpful in understanding the actions and side effects. I’m an MD + computer science hybrid with a background in building interactive tools for helping people understand medical concepts, so I decided to build a free scroll-through explainer in the form of a subway ride. The ride is 7 stops, takes about 5-10 minutes, and doesn’t require an account. This was with the assistance of AI, using multiple tools. The process was far from one-shot and with me heavily in the loop for all the different pieces. I still find it tricky to get something medically accurate and useful when the details matter. I had built a free set of pages, GLP-1 Field Guides, to help people understand the different GLP-1 related drug classes, and I was able to lean on all of that research and sources to get the details right and to provide authoritative sources for the information presented at each subway stop. I had tried building this starting in February but then paused it for a few months because the models weren’t quite at the point where I could make what I had in my mind’s eye. In the final piece, the different animated elements are drawn in code and change with the scrolling. Here are some of the tools I ended up using: - GPT Image and Nano Banana, both via a custom Claude Code skill I wrote, with my own sketches as the base, for the Field Guide visuals that were important for this piece. Nano Banana for the subway graffiti and the organs - Gemini 3.8 Flash TTS for the conductor’s voice - Suno for the song (ride to the end) - Claude Code with Opus 5.5 to build the site and Codex for code review I am curious if this is a useful format for understanding GLP-1s but happy to answer questions about any of the above.
Wishes MCP connects any AI client to your Wishes wishlist with one URL and OAuth. Tell ChatGPT or Claude what you love, and gift ideas land on your list with store links. Share the list, family books gifts, nobody buys twice. Free.
🧠 One shared brain. Your AI, working together. GPT, Claude, Gemini, and more share context, tools, and capabilities across Claude Code, Codex, and the Mobou app. Use Codex features from Claude Code and Claude features from Codex. Your models talk to each other, check each other’s work, and pick up where the others leave off. Free beta for Mac.
Ask Claude, ChatGPT or any AI assistant that supports MCP to build a mind map. It lands in your MindMap AI account as a real, editable map, not a text outline or a static image. Keep editing it in the same chat: expand a branch, change one node, add relationship arrows, switch layouts. Pull meetings or docs from your connected apps into one map, or map a codebase with Claude Code or Codex. No API key. Works on the Free plan with your monthly AI credits.
Homechecker helps buyers and owners understand a home before deciding what to investigate next. Its desktop property reports draw on available records, imagery and optional uploaded documents and photos to highlight potential condition issues, maintenance needs and indicative costs. Built by a chartered surveyor, it also offers free property guides through an MCP connector listed in the Claude directory.
Your ideal customers don't always fit into database filters. Describe what you're looking for in plain English, and Causo investigates public sources to find companies worth pursuing, with evidence behind each match. Dig deeper in Causo's chat or run the same research in ChatGPT and Claude through MCP. Once you know who fits and why, Causo helps find decision-makers, verify their emails, and write outreach grounded in that research. Better GTM intelligence. Less slop.
TLDR: Connect once to your favorite coding agent. Claude Code and Codex find and load relevant skills for the task at hand, from expert developers or your own team. Hey HN I was originally a bit frustrated with having to scour github repos for the right skill while I tried to improve an iOS app's Swift animation. Then I got frustrated with having to force the agent to use the right skill for the right task. So I built Skillstore: a library of engineering skills behind an MCP server. Claude Code and Codex can search it, load relevant instructions, and apply them to your task. You don’t need to know the skill’s name beforehand. Skills are loaded on demand, so you don’t have to keep the whole library in your prompt. (ideally with help you also save tokens and preserve the context window) ex: In our running app, a review using Paul Hudson’s SwiftUI skill caught undersized touch targets and controls that VoiceOver couldn’t meaningfully identify. We fixed those before the release. In our gym app, a Skillstore-guided review caught an onboarding layout overflowing on small iPhones at large text sizes. You describe the work → your agent finds relevant expertise → your agent uses it. You dont have to browse github repos and filter by stars anymore. And you dont have to invoke skills by name You can also give your team a shared library of private skills Been using it inhouse and with friends for the past 2 months. Favorite feature so far: Turning a tutorial into a private skill, so the agent can use that knowledge on future tasks when the time is right You can try the public library without an account https://skillstore-jk.fly.dev/developers#try Send me your feedback :)
Find relevant podcasts, write personalized pitches in your own voice and manage outreach from your email. PodPitch helps PR teams, founders and marketers turn expertise into guest interviews. Public podcast discovery is also available through MCP without a PodPitch account, while existing customers can connect their account for private campaign, draft and inbox tools.
Schedule LinkedIn, X, and TikTok posts yourself or with Claude, Codex, and Cursor via UniPost's hosted HTTP MCP server. Agent posts wait for approval. Get 2 accounts free, unlimited teammates, a full calendar, and available analytics. At 50 accounts: $54/month vs. Buffer Essentials at $195 (72% less). Monthly USD subscriptions; features differ. Facebook Pages and YouTube are in beta; Instagram is coming soon. TikTok posts may stay private during app review. X has a monthly post-credit allowance.
A desktop app for Windows, macOS and Linux that opens, edits and saves PowerPoint files without an account. Edit right on the slide with a PowerPoint-style ribbon, slide thumbnails, a format pane, speaker notes and slide show, and Ctrl+S saves the samepptx. Sign in to create a deck with AI that lands straight in the open presentation, or turn on Connect AI apps so Claude Desktop, Claude Code or Cursor can change slides over MCP.
VberAI is the AI-native production platform for game developers, art teams, and content teams. MCP plugins for Unity, Godot, and Cocos Creator let Claude, Cursor, and Windsurf operate the engine directly. AI Studio ships PSD/Figma designs to the engine and AI Super Matting handles hair-level cutouts in the browser.
Rodz detects 46 types of B2B buying signals in real time: job changes, fundraising, hiring, public tenders, competitor engagement, social mentions and more. For each signal it finds the decision maker you target, enriches email and phone, and pushes the lead to your CRM or outreach tool (HubSpot, Attio, Pipedrive, lemlist, Instantly...) or to your own app by API, webhook or MCP. You pay per lead delivered, with credits, no seats. 150,000+ qualified leads delivered every month.
I work on BlueFacts a stats site for Bluesky, it tracks accounts, custom feeds, tags etc I added MCP support yesterday so you can create your own feeds eg in claude code, or fetch Bluesky stats. We have a real time search so the agent can quickly iterate and refine the feed, exclude unrelated terms etc. It is free, let me know any feedback you have!
Ask ChatGPT or Claude for flights, hotels or seats in plain words, and real prices come back as cards right in the chat. Say "watch it" and SlickTrip keeps watching after the chat ends, then alerts you by email, text or push when the price drops or the seat you want opens up. No dates yet? Pick a route and the months you could go, and we tell you when a good fare shows up. 35 travel tools, free account. 20,000 people joined in the first two weeks, and it hit #1 trending connector on Claude.
🚇 Turn distributed traces, GitHub Actions runs, and coding-agent sessions into animated NYC-subway-style maps: GIF, MP4, or an interactive page you can drop in Slack. Fully local, no telemetry. An agent skill for Claude Code, Codex, and Cursor.
Plan a road trip day by day: add stops, get distance and drive time for each leg, hotels near every overnight stop and attractions along the way. Free for trips up to 3 days. Premium adds unlimited days, link sharing, PNG export and an MCP server.
Make website feedback easy for clients. They point, annotate and comment; your developer team and its AI agents get the same screenshots, exact text requests and replay context. Keep the conversation with the work. Free, Apache-2.0 and self-hosted. Start with one Docker command.
A one-person studio's working notes from shipping real iOS and Android apps, client sites, payments and automations with Claude Code. An App Store and Google Play launch playbook, 105 prompts that actually shipped work, 228 silent-failure fixes (symptom, cause, fix, prevention) and 7 drop-in Claude Code skills. Each guide on its own or all of it in the Builder Bundle.
Deep Code is an open-source AI coding assistant tool deeply optimized for DeepSeek, It supports deep thinking, reasoning effort control, Agent Skills, and MCP integration.It has already gained 2,000 stars on GitHub.It recently launched a Coding Plan, with pricing as low as $9.9/month.
文档： 开源 MCP 服务： 这是一次来自产品方的工具自荐。Sorsa API 提供公开 X（Twitter）数据的只读 REST 接口，包括用户资料、帖子、搜索、关注者与关注列表、列表、社区及互动验证，可用于数据分析、研究和 AI 应用。 使用一个 Sorsa API Key 即可接入，无需提供 X 账号密码或 Cookie，也不需要 X 开发者账号。提供在线 API Playground、OpenAPI 描述文件，以及可在 Cursor、Claude Desktop 等客户端使用的 MCP 服务。 新用户可试用 100 次请求，无需信用卡；更多用量需付费。服务不支持发帖、私信或读取私密账号内容。Sorsa API 是独立服务，与 X Corp. 无隶属关系。
AgentReady is a vendor-neutral, open specification for making websites and products usable by AI agents. It covers discoverability, machine-readable content, authentication, APIs, MCP, WebMCP, and agent payments, with practices grounded in research on real agent behavior.
Bits is a personal agent that runs on your own machine and never sleeps. A small background daemon does the work: routines wake it on a schedule, and tasks keep going when the app is closed. Every step is saved, so a crash just resumes. Memory lives in plain Markdown files you can read and edit. Risky actions pass the Sentinel, a reviewer that only trusts your own words. It answers with live cards and mini-apps, talks by voice, opens with ⌥Space, and works with any model, including local ones.
Voice AI widget for any website. Answers visitor questions out loud, books appointments, takes payments, and live-transfers qualified leads to your sales team. Installed with one line of code, configured entirely through the Talk Kit MCP by AI coding agents like Claude Code.
Most agent testing either traces failures after customers hit them, or makes you hand-write every test and let an LLM guess if it passed. ClientCoded generates adversarial tests and knows the correct answer, scoring against computed ground truth with SQL, not opinion. It catches failures before production, monitors every live conversation with Slack alerts, and turns real production failures into new regression tests automatically. Connect in minutes with one webhook. First audit free.
Doop AI Studio brings AI image and video generation into one developer-friendly platform. Generate cinematic videos with Veo 2 and high-quality images with Nano Banana 2.1 through a web studio, API, and MCP integrations.
Ramorie gives Claude Code, Cursor, Codex and other MCP-compatible agents persistent project memory. Save decisions and bug fixes, retrieve them in later sessions, and inspect saved context with semantic search and a knowledge graph. Memory stays on your Mac; external AI usage is separate. Desktop is free until April 2027, with no account or card required. A paid license is planned afterward. macOS 14+, Apple Silicon and Intel.
Upload a CSV or pick a Google Sheet, and Claude, ChatGPT, Cursor or any MCP client can filter, group and total the rows. The numbers come from a database, not from the model's guess, and every answer says which version of the data it used. An import report flags messy data first. Same data over a REST API for n8n or Make. OAuth sign-in, read-only, hosted in the EU. Free for 3 datasets of 100,000 rows.
RepoGraph AI turns any repository into an AI coding-agent graph — a free VS Code extension. Open a folder: it auto-indexes every file and import into an interactive dependency graph (2D + 3D sphere), tagging each edge EXTRACTED, INFERRED or AMBIGUOUS so you know what to trust. Communities, flow diagrams and Q&A with citations work offline with zero API keys; connect Groq, Ollama, Gemini, Claude or OpenAI for AI summaries, and export REPOGRAPH-CONTEXT.md to ground agents in your real graph.
AtlasMCP lets Claude, ChatGPT, Gemini or Cursor work directly on your WordPress site, so you can stop copy-pasting between chat and wp-admin. Just ask: “draft a post with a featured image”, “which products are low on stock?”, “fix the alt text on my last 20 images”. Connect in one click, choose exactly what the AI can touch, and see everything it did. Works with the plugins you already use, including WooCommerce.
New since the first launch: self-hosted Team features are free with no license key (approvals, SSO, spend-cap holds). Approve risky actions from Slack, Teams, Telegram, Discord, email or the dashboard. Daily money caps and allowed hours per agent, undo for refunds and messages, a one-command install on Linux, macOS and Windows, `agentraas scan` to check an MCP server before you connect it, and keys that can stay in your own Vault or cloud KMS.
TranscribeX records and transcribes Zoom, Google Meet, Teams, Slack and Discord calls directly on your Mac. No bot joins and nothing is uploaded. Whisper, NVIDIA Parakeet, Nemotron and Qwen3 run on-device, with speaker labels, live captions in 30+ languages and AI summaries from a built-in local LLM. It also transcribes any audio or video file, dictates into any app, and has a CLI and MCP server. Free with unlimited transcription; Pro is $39 once, no subscription.
A Discord bot for help channels. When someone asks a question that was already solved, it replies with the past answer, plus an AI-drafted reply on paid plans, before a mod has to. Solved threads are tagged, archived and searchable in Discord, and can publish to a public FAQ site Google can index (opt-in, usernames aliased). Pro and Max expose the archive as an MCP server for Claude, ChatGPT or Cursor. Semantic search on every tier, free plan included. Open source and self-hostable.
A²D is a design-first platform for MCP mocking, API mocking, and agent mocking. Design, mock, validate, and test MCP server specs and A2A agent cards before the real backend exists — then publish to Anypoint Exchange. No code required.
ChatterEase is an affordable AI customer support agent for startups and small businesses. Add one widget to your site to answer customer questions 24/7, capture leads, cite sources, and hand conversations to a human when needed. Train it on your website, docs, and custom Q&As, customize it to your brand, and launch in minutes on WordPress, Shopify, Wix, Webflow, or any site. Supports 50+ languages, analytics, auto-retraining, webhooks, and API access.
Claude Code forgets everything between sessions, and its defaults are not a careful engineer's habits. developer-os is a free, MIT-licensed plugin that fixes both. Your agent searches an Obsidian vault before work and captures decisions into its inbox while it works; you approve what stays. Rules load in every session, skills cover specs, reviews, audits and PRs, and hooks ask before installs or force pushes. Plain Markdown, no server, no telemetry.
BrightSite is a website builder and host that your AI agent can run. Connect Claude, ChatGPT, or any MCP client with one URL and get nearly 300 tools. Ask for a page, a form, a product, or a fix and it lands as a draft under your own permissions. Lumi is built in with no credits. Every plan includes a store, website chat, email campaigns, forms, analytics, and SEO tracking. Unlimited team members. From $29/mo per site, 7 days free, no card.
FreeQrly is a free, privacy-first QR code generator. Static QR codes are generated 100% in the browser — no sign-up, no tracking, no expiry. 16 content types (URL, text, WiFi, vCard, email, SMS, event, Bitcoin, coupons, app links, reviews, MP3, forms, menus & more), batch mode, barcode generator, QR scanner. Unlimited dynamic codes with scan stats, editable after printing — limited-time offer. 10 languages. Free API + MCP server for AI agents. Independent project, not affiliated with FreeQR.
Aerini is an open source, local-first visual workflow engine. Wire nodes on a canvas and run them by hand, on a schedule, or from a webhook, on your machine or on a headless server. No account, no cloud, no telemetry. Extend it with sandboxed WASM plugins, or let AI agents like Claude Code build workflows through the separate aerini-mcp server. Pre-1.0 and still in development, so expect rough edges.
SearchTok makes TikTok and Instagram searchable for your AI agent. Connect it to Claude, ChatGPT, Cursor or Claude Code with one URL and ask what's working in your niche. It finds the outlier posts (ranked by how far they beat their account's usual views), quotes their hooks word for word, and shows the creators behind them. Built for founders, marketers, creators and small agencies.
The Brains turns founders, CEOs and investors into searchable MCP servers built from their public talks, essays, interviews and posts. Ask how Pieter Levels prices, what Paul Graham says about fundraising, or how Ryan Hoover launched Product Hunt, right inside Claude, Cursor, ChatGPT or Codex. Every answer quotes the source with a link. 37 brains today, one URL per brain or one for all. Free, open source (MIT), unofficial and not affiliated with the people.
Convert.Online started as a plain file converter and still is one, across 400+ formats. What's new: an MCP server, so Claude, ChatGPT and Gemini can convert files directly in a conversation, and a REST API with chained jobs, webhooks and a sandbox mode. Also added since the last launch — OCR that makes scanned PDFs searchable in 17 languages, and imports straight from Google Drive, Dropbox and OneDrive.
Track AI infrastructure spending, debt, leases and cash flow with dated sources. Explore the research on the web or query it in Codex and Claude Code with an open-source skill. English and Chinese. No ARGUS API key needed.
Upload your app's real and Agent C designs the full App Store set in about a second: headlines, layouts, 3D device tilt and colors from your app. Edit anything, export at exact App Store Connect sizes, or send it straight into your app's draft version. Also works from Claude Code, Cursor and Codex over MCP. Free to design, $9.99 per finished set, no subscription.
An open-source, self-hosted node that gives your AI agent an address other people's agents can reach. Two people, two nodes, nothing in between: MCP over mTLS, messages sealed end to end, and a per-contact switchboard that decides what each contact may do. Your identity is a keypair you hold, not an account. Add the open-source agent and it answers your contacts for you, asking you only when a decision is yours. One Docker image, SQLite by default, no telemetry. Apache-2.0.
Private, tamper-evident notes for your AI sessions, scripts and agents. Write over MCP, REST or one curl; a hook archives Claude Code and Codex sessions. Every note is Bitcoin-anchored, so you can prove it never changed; and much more, please check it out. Free to start.
OnFlow is a no-code, AI-native platform that allows you to build your own back-office system by defining your own data models and workflows. Everything is accessible and manageable via the UI or via LLMs using MCP. Build your own CRM, ERP, project management tool, HR tool all in one place.
Publish the HTML, Markdown or JSON Claude or Cursor makes. Share a link anyone can open, with pinned comments, versions and private, unlisted or public sharing.HaveTML gives the HTML, Markdown, YAML and JSON your AI writes a permanent link. Publish from Claude, Cursor, Claude Code or Codex over MCP, or drag and drop. Unlisted links open without an account; your team pins comments, and each re-upload adds a version. The first 500 accounts get Pro free for 12 months.
Intelli is now a plugin in ChatGPT’s directory, and works in Claude and any MCP client. Ask in plain language and your AI assistant handles your WhatsApp, Instagram and website conversations: summarise unanswered chats, launch campaigns, build chatbot flows, update contacts and products. 89 tools, OAuth sign-in, no API keys. Nothing is sent, edited or deleted until you confirm. Also new: Instagram/Facebook comment & DM automation.
tldr; Travelers wanted to find and book flights super fast in Claude, so we built this. -- BonBook started as flights via email because I was flying once a week in '24 and the UX got to me. Since then we've handled tens of thousands of requests for hundreds of travelers. The new MCP lets you search live flights with as little/much detail as you like, get options and then book without filling any forms. Points and status intact. All in ~35 seconds.
Spirit Room tests your MCP server, CLI or API agains the jobs users need done. We inspect tool searches,calls, errors and outcomes across models and agent as, then test changes to names, descriptions and resonses. Available through hands-on pilots.
Running multiple coding agents shouldn’t mean becoming their project manager. BAND lets them work together. Connect agents across your existing development tools and give them a shared workspace to coordinate tasks, exchange context, and hand off work without you playing middleman. See what every agent is doing, step in when decisions need your judgment, and scale agentic development across teams. Your agents stay in their existing tools while BAND handles the teamwork.
Notion’s Agents update rebuilds the agent harness so tasks can run for 24 hours or more, with a computer, a browser and a Review for me safety mode. Auto now routes each task to the best model, and Notion says it matches Opus 5.5 and Astra at about 90% lower cost. You can connect a ChatGPT plan, Notion MCP is overhauled, and a new credit system covers all AI usage. Live now for new customers.
I wrote a coding agent that is on par with most coding agents and is particularly well designed for building and running repeatable steps that automate your long running tasks. I experienced many times that skills are not fitted for daily task such data job backfills, CI steps and complex DAGs etc..
ChronoReact is a bilingual (TR/EN) trivia site where Claude, Gemini and Groq autonomously research, write, and fact-check a new card every day — no human writes the content. It also has a daily logic/math puzzle, a countries leaderboard, and its own MCP server for AI assistants to query.
Arva SEO helps business owners find their next SEO step. Audit your public website, research keywords, plan and improve content, and track progress in one workspace. Works across website platforms, with an optional WordPress Connector for health checks, SEO titles, meta descriptions and drafts. Connect Claude, ChatGPT, Gemini or Grok through MCP, or invite a specialist to selected websites. Start free with your first website.
Give your AI agent financial rails for paid ads. Issue virtual cards with custom spend caps, fund cash gaps so winning campaigns keep scaling while revenue catches up, and let your agent check charges against live ad results via MCP. Built for businesses that scale on paid ads.
We’re launching Kimpa, a security-first, privacy-focused business email and workspace for modern businesses. Mail, calendars, contacts, documents, spreadsheets, presentations and team collaboration come together in one connected platform. MCP support lets teams connect compatible AI assistants and agents to their tools while keeping greater control over business information. We’d love your feedback on building a more secure, private way to work.
A Claude Code plugin that sends hook events (task finished, waiting on your input, a tool you care about was used) to Slack, Discord, email, WhatsApp, desktop voice and web push. Walk away from long runs without missing the moment.
When an AI coding agent hits an obscure error, it burns thousands of tokens retrying dead ends. Myrmo is an open-source shared memory layer. Agents publish verified fixes as "trails". The next agent hitting the same error finds the fix instantly. Stop repeating mistakes.
DSHX is a build-time toolchain for out-of-tree DeepSeek Harness plugins. It gives you typed Host/Client authoring, a bounded Vite build, offline diagnostics, Profile-aware development and reproducible templates, while the official DSH/Cordis services keep ownership of execution, registries, transport, HMR and disposal. Run check in CI, build Host and Client artifacts with declarations, and keep plugins in sync with the runtime you test against.
I made Agentless server management GUI over plain SSH. In Litedeck, we can show and download file, service, docker container, terminal and security, and resource monitoring. Because, I have some Raspberry pi4, mini pc(N97, N100), jetson nano but cockpit or other app install in the server, but I don't want that. My servers haven't lot of resource. so I made litedeck. And Claude or Codex install in server is heavy too. AI can connect to ssh but, I want to watch what AI's do. So I make to tracking AI and show graph view using my Litedeck mcp.
Glyph is MCP-native code intelligence for AI coding agents. Connect a GitHub and agents get hybrid semantic, keyword, symbol, and graph search not just snippets, but definitions, callers, blast radius, test plans, safe change order, and a verification receipt before they say done. It works with any MCP client, keeps diffs local, and turns blind edits into reviewed, evidence-backed changes.
Anja is the engineering system one developer built around Claude Code while shipping a production web app between March and October 2026. It is extracted and cleaned for others to study, use and adapt. Claude Code supplies sub-agents, skills, hooks and memory. Anja adds the discipline between them: who delegates to whom, how sessions share one branch, who reviews what, how a session recovers after a compaction, and how a correction becomes a lesson, a rule, then a guard. Open source, Apache 2.0.
What’s up HN?! I'm Benji, and I’ve been building Frink since ~April this year. Frink is a desktop-first, open-source (MIT) app for running coding agents in parallel and multi-step workflows that start on their own. It runs on your machine (macOS and Linux) with your own provider login. It currently supports Claude Code, Codex, Cursor, Copilot, and OpenCode, but the latter 3 are in preview. No Frink account to use the app. The biggest USP of Frink is workflows: think n8n for agents. You describe a routine in a sentence, for example: "every weekday at 9am, run the tests; if any fail, have an agent fix them, then wait for my approval before it opens a PR." Then, Frink builds it as a workflow (or simply, Flow): a list of steps you can read and change before it runs. A Flow can start on a schedule, from any webhook, or from an event from a third-party such as Linear, Shortcut, ClickUp, Sentry, PostHog, etc. Steps can be agent turns, shell commands, conditions, approval steps that stop and wait for you, or steps you write yourself in code (or ask the agent to write one). Everything that needs you (an approval, an agent's question, a finished task to review) lands in one Work Queue. Other stuff: * Up to four agents side by side, each on its own git worktree if you want. Or, many agents out of direct view. * Every change shows as a diff; there's a code editor, a terminal, etc. * Local-first: chats, Flows and tasks live in SQLite on your machine. * Easier provider parity: currently working on and improving the tooling setup between providers. Once inside Frink, irrespective of what provider you use, the hooks, MCPs, and other tooling work out of the box between providers. Why I built it: I started building it when there weren’t many other competitors; now the industry is swamped with ADEs and harnesses. However, in my opinion, each still lacks good automation for more fine-tuned scenarios. I doubled down on the flows to help me with my actual day-to-day job as a SWE, and over time it’s helped me automate most of the PRs being shipped across my personal projects, team, etc., to the point that it’s a continuous loop and helps a whole ton. Notes: Windows isn't supported yet. Flows only run on Claude Code and Codex for now. It's a solo project, so expect sharp edges.
Blob0 brings Cloudflare R2 and S3-compatible storage to your Mac. Browse, organize and preview files, paste uploads from your clipboard, and share public or expiring links. Connect with Cloudflare, track usage in a visual dashboard, and turn your buckets into an image host for PicGo-compatible editors. Local image compression, dark mode, a CLI and Agent Skills complete the workflow. Your files stay in your own buckets.
Create a MacOS mcp bridge for agents to control app reminders, more functions than similar projects. see https://github.com/xiaofeng-han/apple-reminders-mcp for details
Your AI agents (Claude Code, Codex, Cursor, claude.ai, ChatGPT) publish HTML, Markdown, diagrams and PDFs over MCP. You review them in a private, passkey-protected gallery with version history and live reload, comment on them for the agent to fix, and share links that expire. One ~10 MB Go binary with automatic HTTPS, Docker, or the free Cloudflare Workers edition. MIT licensed.
PrivOS is the all-in-one AI-native operating system for enterprise work. Chat, Docs, Kanban Tasks (lists with table and Kanban views) and tailored MCP apps live in one place, and AI agents act on them: @-mention a list, file, folder or app in a room and the agent works on it from its own sandboxed container. Self-host the Community Edition free for up to 10 users (beta) or use our Cloud, hosted in the EU. Bring any Anthropic- or OpenAI-compatible model, including local ones.
Your AI agent publishes a page, report or folder of docs (HTML, Markdown, PDF, Word, Excel) to a real link. Invite your client by email; they pin comments on the page; the agent reads them over MCP, fixes, republishes and replies. Unlimited sites, free.
Swiss cloud storage and secure file sharing for indiiduals and teams. Files hosted in Switzerland, privae vault, expiring and password-protected share links free 1 GB plan. Includes an MCP server for AI assisants.
Currently available in Argentina only. UNAI helps small businesses build an AI-designed website and run their store, bookings, CRM and finances in one place. Accept payments with Mercado Pago, manage products and appointments, and let an AI agent work with your business data. Connect through MCP to manage your business and edit your website from compatible assistants like ChatGPT and Claude. Start with a 15-day free trial, no credit card required.
Jellyfish is an AI browser agent for hard-to-fill roles. Give it a job and hiring context. It searches practice websites, research papers, directories, company pages, and the wider web, then checks candidates against the requirements. Get a shortlist with evidence and source links explaining each match. Built for recruiters looking beyond the usual candidate databases.
Screener MCP is an open-source Model Context Protocol server that gives Claude live access to Indian equity market data from Screener.in, NSE, and MCX — company research, stock screening, financials, portfolio tracking, and more.
AewolFlow is an AI research workspace for literature reviews. Start with a question, choose your papers, compare methods and findings, and draft from selected evidence. Keep quotations and source links beside your work so you can check claims in context. Try one comparison of 2–3 arXiv papers free; Pro adds literature reviews, PDF reading, and connections to Claude, Codex, and Cursor.
Turn your logo into a cast of paper sticker characters. Open-source Codex skill with 15 real examples, original logos, exact prompts and transparent PNGs. Bring an image-generation tool; provider costs apply.
Muze AI connects Chrome to Claude and other AI assistants through MCP, so they can work with your actual browser—not just talk about it. Let your AI inspect and switch tabs, navigate websites, read page content, extract information, interact with forms, and run browser workflows. Fine-grained permissions and one-time approval for sensitive actions help keep you in control. Built for people who want AI to do useful browser work, not just answer questions.
ActOnce Spend Gateway helps you give AI agents spending power while keeping control. Set purchase limits, allow trusted merchants, block categories and flag duplicate requests. Purchases that need your input go into an approval queue. It works with MCP-compatible agents, including OpenClaw and Hermes, and never handles payment credentials. Start with the free, MIT-licensed local gateway, or choose a paid plan for hosted dashboards and team controls.
Every protocol you try is an n=1 experiment, and your DNA is the one variable that never changes between runs. Helisoma reads your raw DNA file in your browser, finds the markers for sleep, caffeine, training, mood and longevity and serves them as an MCP server for your AI, so it can finally see the whole picture.
Check Great Britain goods vehicle and PSV operator licences and see changes between published DVSA files. Free lookup pages bring licence, company and regulatory context together with source dates. Developers get a free JSON API and 15 read-only MCP tools, with no account or API key. Optional paid monitoring adds alerts. Built by Luli Labs Limited; independent of DVSA and the Traffic Commissioners.
Most sites fire Google Analytics, GTM and the Meta Pixel in the first 1.5 seconds, before anyone touches the cookie banner. Regulators look at that. CookieInspector 2.0 loads your site as a first-time visitor and shows every cookie, script and tracker that fires before consent, and the tag behind it. New: an MCP server for ChatGPT, Claude and Cursor, scans from the US or Germany, a How-to-fix guide, Consent Mode v2, CCPA/CPRA, multi-page scans and monitoring. Nothing to install. Free scan.
Hi HN, We (Polytomic) noticed that LLM agents (Claude, Codex, Grok, Amp, etc) are inconsistent in their MCP implementations. We made tables inspired by the browser-compatibility tables of yore that list the differences. Hope you find this as useful as we did!
Run Claude Code, Codex and Cursor CLI side by side. Kith adds scoped work claims, isolated Git worktrees, shared port reservations and supervised command locks through MCP. Check the combined result with your project's verification command before integration. Failed checks keep the target unchanged and preserve the worktree. Free, open source and local—no hosted service required.
Scribe turns every meeting into one searchable record. It joins Zoom, Meet and Teams and records in-person conversations in multiple languages. Ask across your meeting history for decisions and action items in seconds. Tasks land in Linear, Jira or Asana with the sentence that created them. MCP tools let Claude Code and Cursor act on your meetings. Fully white label. Priced per meeting hour, not per user. You own your transcripts, so you can reuse them anywhere. Run a trial for 15 days.
OQRO reads SEC Forms 4 and 144, 13F fund filings, Congress trade reports, 8-K events and federal contracts, and explains each filing in plain words with the link to the official document. Each insider purchase is compared with the company's own filings since 2006, so a rare one stands out: first buy in years, largest ever, several insiders the same week. Plus a written context on each filing, baskets against the S&P 500, alerts, an assistant that cites every row, and an MCP server. Not advice.
video2ctx makes videos queryable for AI agents. Search and inspect videos, get answers cited to the second, monitor channels, and save moments into projects. Use it through agent skills, CLI, API, npm package, or try our video agent directly. Open source with 1,000 free credits for everyone.
I built a claude mod that searches every session (across harnesses) on each prompt. On every prompt, it runs a series of decision models to pick the right sessions + messages and adds anything relevant to the context. Requires no indexing or setup and works out of the box https://youtu.be/o5vEf8tgus4
Record a meeting or write your day, and secondbrain turns it into the recap, tasks with an owner and a date, a planned week and a living wiki of your people, projects and decisions. Connect it to Claude or ChatGPT through MCP and they can search and write to it. Web, Mac, Windows, iOS, Android. Free plan with no time limit.
Hello HN! Today I am presenting you with Peras, the tool which governs your Claude Code and Codex usage, and enforces limits. I built it because I got tired of constantly having to spam /usage to check my usage, and decide if a task is worth risking the agent running out mid-task. Peras enforces a limit with 2 main tiers. At 75% weekly usage you will get a notification, and at 90% Peras will automatically pause the agent. These limits are the default and they are customizable to your own liking. You may ask "how does it know my usage in Claude Code or Codex?" Well, it's pretty simple. Peras gets the Claude Code usage % from the status line, and Codex's one comes from its local app-server. The "pause" Peras executes is a hook that refuses the agent's tool calls and ends its turn. A new feature I implemented is one that makes agents aware of their own usage. When you send your prompt, Peras attaches the usage %. It sends it at the start of the session, then again at 50, 75, 90 and 100%, or if a reset is coming in 30 minutes, so the agent can hurry up and actually capitalize on the remaining usage, if any. Just to clarify, this feature is informational and the pause is what enforces the limit. Before I switched to native SwiftUI, I used SwiftBar. The problem was that for the mascots patrolling their usage bars, I had to re-encode every menu image 8 times a second! There was also stuff SwiftBar wasn't capable of, like the menu being able to appear over fullscreen apps, pinning it as a floating panel, etc. What is the difference between the free and paid version? Well, it is MIT, first of all. The CLI, the hooks, the SwiftBar menu, and the app's source are all on GitHub, but building the app needs macOS 13+, Node 18+, "xcode-select --install" and then finally "npm run build:app". The €4.99 buys you a signed copy which is notarized by Apple, with updates included. The guard works with no menu at all. To install, just run: npx peras-cli setup npx peras-cli codex setup And now we get to the limits part. For now it's subscription plans only, and when running "claude -p", the pause uses the last interactive reading, so it can come late. Also I feel I have to mention that the guard rail is not a security boundary. It runs inside the agent. The menu is macOS only. The CLI and hooks pass tests on Linux in CI, but I don't use them there day-to-day. Last thing before I end this post, I would really appreciate feedback from you, whether install worked for Codex users, whether you have encountered any bugs, or whatever you feel like sharing.
Zavik is an open-source dev CLI & AI-orchestration engine. One call detects your stack and briefs any AI coding agent — runbook, entrypoints, trust (zavik context --json). Another audits your dependency supply chain and gates PRs via a GitHub Action with SARIF output. Plain CLI: stable --json and exit codes are the interface — works with Claude Code, Codex, Cursor or no agent at all. Free forever (MIT core); the paid add-on is a daily signed risk dataset, never locked features.
AURA GEO checks what keeps AI search from seeing your site (robots.txt, HTML, sitemap, indexing), tracks real referrals from ChatGPT, Gemini & 9 more AI services, and hands fixes to your own agent via MCP. Submit updates via IndexNow. Free plan for 1 site.
Look up health and ownership for 48,000+ MCP servers, or audit your whole MCP config in one call. Every claim links to public evidence. Free 10/day. REST and MCP for agents.
PromptingEasy is a free, vendor-neutral website for learning to write better prompts for ChatGPT, Claude, Gemini and other AI assistants. It offers step-by-step tutorials from beginner level to how LLMs work (context, RAG, agents), ready-to-use prompt examples, prompt generators for writing, email, CV, studying, small business and translation, practical use cases, and answer-quality tests using the CLEAR method. Available in 18 languages, no account needed.
Cross-client AI memory hub with bi-temporal governance, supersession chains, hash-chain audit, and 9 MCP tools. One SQLite shared by all agents via file-level pointers. Local-first, zero cloud.
Turn Linux servers into one Docker cluster. Deploy and scale apps, manage workers and encrypted networking, and monitor workloads in a web dashboard. Connect ChatGPT or Claude through MCP. Built in Go by NEXUS AI. Apache 2.0 licensed.
Hi HN! We're releasing an open-source CLI for hillclimbing Agent Experience. That means optimizing your agent-facing surfaces (Skills, CLIs, SDKs, and docs) so that your users' coding agents can easily use your product. Pointing your own agent at the is enough to get started. https://github.com/gauge-sh/gauge-cli We've found that power users generally want to hand a CLI/MCP to their agents and get started from the terminal. They rarely want to log in and use the web UI themselves. We think developer tools must be agent-friendly to remain relevant. Here are some common findings that Gauge identifies: - agents get confused with CLI/SDK behavior and start source diving - interactive commands block agents - agents use deprecated/old CLI commands - agents don't use Skills when you'd expect - agents confidently assert limitations that don't exist - inconsistent terminology confuses agents - agents can't figure out how commands fit together into a workflow Locally, the Gauge CLI uses Docker containers to run agent harnesses (you can bring your own subscriptions). It finds where agents get confused or stuck and tests changes using your pre-release artifacts (Skills/CLI etc). We normalize events from Claude Code, Codex, Pi and OpenCode into standardized transcripts, and store them locally, alongside structured information in SQLite so that your agent can easily analyze the data. This enables automatic hillclimbing toward a great Agent Experience. We also offer hosted compute that enables more features, built on our custom Gateway (for credential brokering and transparent, per-session response rewriting). Hosted compute enables many more concurrent sessions, easier collaboration with teammates, a GitHub App for PR reviews, sessions with logged-in MCPs/APIs, and most importantly it lets you close your laptop while the sessions actually run. You don't need a Gauge account to run the full loop end-to-end though. We expect that local execution with the CLI is enough to find and fix several agent experience issues in a typical project. You'll be able to validate fixes for these issues locally, before publishing new versions (we inject pre-release Skills and code artifacts into the sessions). We would love your feedback!
Running several products means scattered email tools and bills that grow per seat, per inbox and per subscriber. MarkRelay puts it all in one app on your own Cloudflare account, deployed in one click: - Webmail for every domain you own - Broadcasts with a visual template designer - HTTP API + webhooks for products and AI agents - Per-mailbox access for your team Your data stays in your account. One flat license, no usage-based pricing.
PaperPilot is an open-source, privacy-first PDF engine built in memory-safe Rust. It runs 100% locally via WebAssembly in your browser and native desktop—zero bytes ever leave your machine. Features: • Split, merge, compress, watermark, crop & encrypt • Visual multi-step pipeline builder • Embeddable JS automation widget • Native MCP server for AI agents • 100% free and open-source under MIT
Advancer.World is for organizations that strengthen many partners, members or grantees along a shared path: NGOs with local partners, federations with national members, foundations with grantee cohorts. Put the program on one grid of maturity stages by topic, see where every partner stands, and attach the work each one owes: questionnaires, documents, initiatives, KPIs. Partners work in 22 languages and never pay. Ask Claude or ChatGPT about it through MCP. Free to start.
Cachejay keeps track of your workday on your Mac (screen text, calls, coding-agent sessions, files and to-dos), files it by project, and gives that context to ChatGPT, Claude, Cursor and Codex.
QingIcon is a free, open-source icon library with 3900+ icons across 34 categories. A single SVG source powers four styles — line, solid, adjustable stroke, and a live hand-drawn sketch. Ships as React/Vue components (tree-shaking friendly, currentColor theming), a Figma plugin to drag icons onto the canvas, and an MCP server so AI tools like Claude and Cursor can use icons directly.
TicketGlide brings attraction tickets, tours, day trips and practical guides into one easy-to-use travel comparison platform. Explore in four languages on desktop or mobile, with no app to install. Connect an AI assistant through our public MCP endpoint without an account or API key to search guides, read guide details and create travel search links. MCP is read-only; booking and payment happen on an external provider website.
Know who’s on your website before login and keep identity, permission, memory, and context connected across digital, physical, and agentic interactions.
22 focused audits for Swift and SwiftUI apps. Fill in your project once and every prompt updates. Copy an audit into a fresh Claude Code session with a Swift agent skill, then mark it done once you've triaged what it finds. They cover concurrency, security, architecture, performance, accessibility, Liquid Glass, testing, interface copy and more. Most ask for findings by file and line, with a fix for each. Free and open source (MIT). Your project details stay in your browser.
Open-source agent skills that turn Claude Code, Cursor or Codex into an AI CMO. Give your agent your website and it learns your brand. It studies your competitors' ads, ranks what you should test next, writes the copy for each platform, and makes the ads (UGC videos, product ads, cartoon ads, image ads, product photos and more). It checks in with you at every step, and nothing expensive renders until you approve. Bring your own model keys, or pay per use with ours. Apache-2.0.
Hubzoid is an open-source framework for internal AI agents. Put your team’s knowledge, rules and tools in one shared Hub, then use it in chat, workflows and compatible assistants via MCP. Install: pip install hubzoid. Self-host it. Apache 2.0, team controls included.
Live on-chain data for Base Mainnet — gas, DEX arbitrage, whale tracking, Aave flash loans & liquidation, token security, contract profiling. Pay per call in USDC via x402. No API key, no subscription. MCP server included.
Overwhelmed and disorganized by Codex and Claude Code? Nimbalyst wiki is a local wiki your agents write for you so you can think, organize, and set direction Your agents organize decisions, open questions, and the big picture into a wiki you can edit visually. It is a free plugin for Claude Code, Codex, Cowork, Cursor, VS Code or any MCP agent. No app or account needed. Pages are plain markdown, drawing, csv, clickable html files in your repo, and Claude opens them in your browser for editing.
