Category
633 posts in AI Setup
Setup notes for local LLMs and coding agents. What to install, what to configure, what to skip.
cmux is a Ghostty-based macOS terminal with vertical tabs, notifications, and programmability aimed at running multiple AI coding agents at once.
agency-agents is an MIT licensed, 157,285 star repository of specialized AI agent personas, each one shipping its own process and deliverables.
gstack is an MIT licensed TypeScript toolkit of 23 Claude Code tools covering CEO, design, engineering management, releases, docs and QA.
pstack-claude ports Poteto's pstack agent workflows to Claude Code, Codex, Pi, OpenCode and Gemini as an MIT licensed JavaScript project.
OpenMontage is an AGPL-3.0 Python system claiming 12 video production pipelines, 100+ tools and 700+ agent skill files, with 63,249 GitHub stars.
text-to-cad is an MIT licensed Python project with 16,887 stars that gives AI agents CAD tooling for mechanical engineering and robotics work.
cloudflare-os is an agent workspace built on Cloudflare Workers for creating documents, building apps, and running agents with company context, with 10,581 GitHub stars.
claude-mem keeps context across AI coding agent sessions, compressing session activity and injecting relevant memory back, with 95,605 GitHub stars.
codegraph indexes a local code knowledge graph, auto syncs on code changes, and names nine agent harnesses as targets. C, MIT licensed, 72,972 stars.
google/skills collects Agent Skills for Google products and technologies. Apache-2.0 licensed, Python, 20,758 stars, 1,723 forks, pushed yesterday.
marketingskills packages CRO, copywriting, SEO, and analytics guidance as skills for Claude Code, Codex, and other AI agents. MIT licensed, 52,424 stars.
caveman is a Go skill and proxy that claims to cut coding agent token usage by 65%. Apache-2.0 licensed, 109,113 stars, 6,316 forks.
UniMate is a SIGGRAPH Asia 2026 research project animating varied skeleton types with a single unified model. MIT licensed, 1,086 stars.
impeccable packages design guidance into a form AI coding harnesses can act on. Apache-2.0 licensed, 73,704 stars, pushed daily.
HeyGen's framework for turning HTML, GSAP animation, and FFmpeg into rendered video, pitched at agents rather than editors. 55,364 stars.
context-mode sandboxes tool output, persists session memory, and enforces routing across 17 agent platforms through MCP and hooks. 24,793 stars.
Cursor's plugin specification and its official plugins, written in TypeScript. 9,327 stars, 883 forks, and no license file on the repository.
mattpocock/skills collects the agent skills the author ships from his own .agents directory. 273,938 stars, 23,000 forks, MIT licensed.
PageIndex is an MIT-licensed Python tool that indexes documents for reasoning-based RAG, replacing vector similarity search with an LLM reading a document tree.
NVIDIA OpenShell is an Apache-2.0 Rust runtime that the company positions as a safe, private place to run autonomous AI agents.
Mobile MCP is the Apache-2.0 TypeScript Model Context Protocol server for mobile automation and scraping across iOS, Android, emulators, simulators, and real devices.
Claude Code Action is the MIT-licensed TypeScript GitHub Action published by Anthropic. Stars stand at 9.1K with 805 open issues since May 2025.
NVIDIA Model-Optimizer is an Apache-2.0 Python library that unifies quantization, pruning, distillation, NAS, and speculative decoding for TensorRT-LLM, TensorRT, and vLLM.
Hindsight is a Python, MIT-licensed open-source memory layer for agentic AI. Tracked on GitHub at 32K stars with topics covering agent memory.
Paperclip is the open-source TypeScript app teams use to manage AI agents at work. 87K stars on GitHub, MIT licensed.
Supermemory is a fast, scalable memory and context engine for AI apps, runnable fully locally or on Cloudflare. TypeScript, MIT, 30,303 stars, docs at supermemory.ai.
OpenSpec brings spec-driven development to AI coding assistants like Claude Code and Cursor. TypeScript, MIT, 69,373 stars, with docs at openspec.dev.
WeKnora is Tencent's open-source LLM knowledge platform for turning raw documents into a queryable RAG, reasoning agent, and self-maintaining Wiki. Go, 25k stars trending.
Voicebox is a TypeScript AI voice studio for cloning, dictation, and creation. MIT-licensed, 54k stars, runs CUDA, MLX, Whisper, and Qwen3-TTS under the hood.
Anthropic ships an open-source repository of plugins for Claude Cowork aimed at knowledge workers. Apache-2.0 Python repo, 24k stars, the canonical Anthropic-curated set.
ECC is an MIT-licensed agent harness performance system bundling skills, instincts, memory, security, and research-first development for Claude Code, Codex, and Opencode. 260k stars.
Cline is an Apache-2.0 autonomous coding agent that ships as an SDK, CLI, and IDE extension. TypeScript, 68k stars, the long-running open alternative to closed coding agents.
Claude Code is Anthropic's official agentic coding tool that lives in your terminal and executes routine engineering tasks. TypeScript, 145k stars, the canonical Claude CLI.
Pi is a TypeScript AI agent toolkit with a unified LLM API, agent loop, terminal UI, and coding-agent CLI. MIT-licensed, used by 100k+ stars worth of developers.
LibreChat is a self-hostable AI chat UI that supports OpenAI, Anthropic, AWS, Azure, DeepSeek, Gemini, and more. TypeScript, MIT-licensed, with MCP, agents, and multi-user auth.
Addy Osmani's agent-skills is a JavaScript-based collection of production-grade engineering skills for AI coding agents, with first-class support for Claude Code, Codex, Cursor, and Antigravity.
VoxCPM is a tokenizer-free text-to-speech model from OpenBMB that handles multilingual synthesis, creative voice design, and zero-shot voice cloning in a single architecture. Apache 2.0.
VoiceStudio is a fully-local, open-source alternative to ElevenLabs: voice cloning, voice design, video dubbing, dictation, transcription, and audiobook creation across 646 languages. AGPL-3.0.
TradingAgents models a trading firm as a multi-agent LLM system. Fundamental, sentiment, and technical analysts debate trades while risk and portfolio managers oversee execution. Apache 2.0.
MiroFish is an open-source swarm intelligence engine that builds knowledge graphs and uses multi-agent simulation to forecast everything from financial markets to geopolitical events. AGPL-3.0.
Agent-Reach gives your AI agent a single CLI to read and search Twitter, Reddit, YouTube, GitHub, Bilibili, and XiaoHongShu — no API keys, no per-platform scrapers to maintain.
Open-source Python framework for frontier music generation with symbolic planning, zero-shot covers, style transfer, and agentic music editing.
A curated collection of extracted system prompts from Claude, ChatGPT, Gemini, Grok, Cursor, and other major LLMs. Updated regularly, CC0-licensed.
A curated collection of 100+ open-source AI agent, agent skill, and RAG app examples in Python. Apache 2.0 licensed and actively maintained.
Official GitHub toolkit for spec-driven development with AI agents. Turn a high-level intent into a working spec, plan, and implementation loop. MIT licensed, Python.
Open source research knowledge base where AI agents collect, search, and synthesize web research into a persistent, searchable corpus. Claude Code skills ready.
Open source AI agent for mathematical modeling competitions. Ingests a problem, plans an approach, runs the analysis, and produces a submission-ready paper.
Open source skills framework and software development methodology for coding agents. Works with Claude Code and other agentic CLIs. Shell-based, MIT licensed.
Autonomous AI trading agent for Polymarket, Kalshi, Binance, and Hyperliquid. Open source TypeScript stack with Claude under the hood.
Open source desktop wrapper around an AI coding agent. Electron UI, Rust host core, pi Agent Harness, and user-installable plugins for local-first coding workflows.
Claude Code skill that forces your AI agent to give direct, scannable, ADHD-friendly answers instead of burying the punchline under walls of text.
Score every MCP agent run for PII leaks, prompt injection, hallucinations, and cost overruns. Zero SDK, MIT licensed, self-hosted dashboard in one npx command.
REST API and MCP server that lets any AI agent search for and retrieve agent skills at runtime. Aggregates from skills.sh, GitHub, and community sources.
Run Claude Code, Codex CLI, and Gemini CLI from your phone or tablet. CloudForge brings a full terminal to any device with web-based access and BYOS server support.
SubVault is an open-source MCP server that gives Claude, Cursor, and Copilot a shared vault. Say vault to save decisions, remember to load them back — one memory layer for every MCP client.
Sonarly connects to Sentry, Datadog, and Slack to auto-triage production alerts, find root causes, and suggest fixes through an AI agent. Built by YC W26.
Mireye provides AI agents with a single API and MCP server for physical world data — elevation, flood zones, parcel records, land use, and more — with a citation attached to every field.
DroidClaw hooks an LLM brain onto ADB so you can describe a goal in plain English and watch your retired Android phone tap, type, and swipe its way through any app.
A secure persistent personal agent server written in Rust — sandboxed tool execution, multi-provider LLMs, voice I/O, SQLite plus vector memory, and built-in integrations for Telegram, Discord
A free interactive timeline that maps every major large language model from the original Transformer (2017) through GPT-5.3 and beyond. Includes Claude, Gemini, LLaMA, Mistral, and DeepSeek with
WinGPT ports an AI assistant to Windows 3.1 using Open Watcom, WolfSSL, and the OpenAI API — running natively on a 386 with under 1 MB RAM.
Telem AI provides a unified API for AI agents to route web searches and page fetches across multiple providers, with automatic failover and per-request observability.
A free browser tool from Vercel Labs to chat with GPT, Claude, Gemini, Llama, and Mistral simultaneously. Compare responses in real time, no account needed.
Almanac is a YC S26 AI agent that maintains a shared company wiki, connects to Slack, iMessage, Gmail, GitHub, and more, then acts on your behalf. From $30 per user per month.
A cross-platform, local-first Markdown notebook with built-in local AI models for text, voice, OCR, and image understanding — no cloud required.
OzBrain connects Claude, ChatGPT, and other AI agents to a shared knowledge base they can all read from and write to, eliminating duplicated memory across platforms.
Run Claude Code, Codex, and other coding agents on your own infrastructure with OpenHands Agent Canvas – a self-hosted control center for AI-driven development.
Open-source drivers, MCP server, sandbox, and benchmark for building computer-use AI agents across macOS, Windows, and Linux.
Open-source N-L-to-SQL engine that routes plain English queries against PostgreSQL, SQLite, BigQuery, and more. Deploy as a Docker container or integrate directly into your app.
All-in-one job search workspace with AI resume building, cover letter generation, and one-click applications.
Cyberdesk turns natural-language instructions into reliable, self-correcting Windows desktop agents. Built by YC S25, it handles legacy apps in healthcare, finance, and logistics without brittle RPA
GodModeHQ lets sales teams run AI agents as spreadsheet columns to research prospect companies, uncover pain points, and craft personalized outreach without spray-and-pray.
A 125M-parameter transformer runs locally on iPhone 15 and autocompletes piano performances in real time at 108 notes/sec — no cloud required. GitHub Copilot logic, but for MIDI.
mcp-use is an MIT-licensed TypeScript framework for building MCP servers and AI apps with Zod schemas, React Views, and built-in debugging tools.
TinyFish runs hundreds of browser sessions simultaneously with a serverless architecture, scoring 82% on Mind2Web hard tasks against OpenAI Operator 43%.
Hopper is an agentic TN3270 emulator that lets AI agents navigate ISPF, write JCL, debug jobs, and query VSAM inside z/OS — built by Hypercubic, a YC-backed startup.
A privacy-first AI overlay that floats over any app, transcribes meetings live, and answers questions on-screen without switching tabs. Built with Tauri, just 10MB.
Zenflow replaces ad-hoc AI prompting with spec-driven workflows, RED/GREEN/VERIFY loops, multi-agent coordination, and automated code verification before human review.
Langy is an AI coding agent inside LangWatch that reads production traces, writes scenario tests, opens pull requests, and proves fixes in your CI pipeline.
Agent-first web browser for AI agents. Simulates human typing and mouse movement. Built on Playwright with a CLI for profile and browser context management.
Open-source agent that runs local AI models out of the box with no setup, API keys, or cloud dependency. Browser automation included via skills.
One API to route across top AI models with zero cold starts, per-second billing, and an OpenAI-compatible endpoint. Built on NVIDIA GH200 hardware.
A local-first, open-source voice note-taking app with semantic search, auto-tagging, and a Rust CLI for AI agents. Notes stored as plain markdown, searchable by meaning.
Run LLMs locally on iOS and Android with an OpenAI-compatible API layer, custom quantization kernels, and zero-copy edge-cloud handoff.
Workser is an all-in-one AI platform: build web apps by describing them, deploy AI agents with persistent memory across 100+ connected tools, and automate workflows — no server management needed.
A secure persistent personal AI agent server written in Rust. One binary, sandboxed execution, multi-provider LLMs, voice, memory, and MCP tools.
A CLI that gives AI agents their own persistent cloud VMs, from 1vCPU to 60 vCPU plus 8xH100 GPUs, billed by the minute.
Bullet routes straightforward coding tasks to fast models, searches only relevant files, and runs independent tool calls in parallel — cutting agentic coding time with a leaner orchestration layer.
Memex is a cross-platform desktop app for vibe coding that combines a chat interface with a coding agent, built with Rust, Tauri, and bundled Python.
UpTrain is an open-source platform with 20+ pre-configured LLM evaluation checks, root cause analysis, and a local dashboard. Apache-2.0 licensed.
Moltis is a secure, persistent personal agent server written in Rust. One binary, sandboxed execution, multi-provider LLMs, voice, memory, and built-in integrations for Telegram, Discord, Slack
Sentrial monitors AI agent sessions, automatically detects failures, diagnoses root causes with AI, and generates code fixes via GitHub integration.
CraftBot is an open-source AI agent that builds, evolves, and operates its own SaaS tools while working alongside you. Self-hosted, MIT licensed, with 40-plus agent profiles and 120 automation
Kolosal AI is an open-source C++ desktop app for running LLMs offline on any AVX2 CPU or AMD/NVIDIA GPU. ~20MB compiled, Apache-2.0 licensed.
Stop re-explaining your project to Claude every session. ContextForge adds persistent memory to Claude Code, Cursor and Copilot via MCP.
Insurance for AI agents, humanoid robots, and drones with audit-gated underwriting and minute-level quotes.
Julie is an open-source desktop AI overlay for macOS and Windows that reads your screen, responds via voice or text, and runs browser and computer automation agents using Groq-powered LLMs.
Sx 2.0 turns any shared folder into a skill server for AI agents. No server, no git, no account needed - just point Dropbox at your team vault and share skills instantly.
Open-source context retrieval layer connecting apps, databases, and documents to AI agents via unified search, SDKs, and MCP — 50+ integrations, self-hostable.
AI reading tool that analyzes complex PDFs block-by-block with real-time explanations, adapting to your learning goals.
Give AI agents structured access to any macOS app via the accessibility tree—no screenshots, no pixel matching, just 78–96% token savings vs full-tree dumps.
Open-source browser automation tool by Selenium's creator — gives AI agents a verification layer through CLI, MCP server, or JS/Python/Java libraries.
Every speech model benchmarked across 10 languages, routed through a single OpenAI-compatible API. Built by YC S26.
Superpowered captures meeting audio at the OS level, generating AI notes, action items, and summaries without a bot joining your call. Free tier available.
Deft Writing uses distribution fine-tuning to produce AI-generated text that reads like actual human writing instead of generic LLM output.
Self-hosted issue tracker for coding agents. One ~25MB binary, SQLite, native MCP across 11 AI clients. Plans and issues outlive the session.
One API to scrape and crawl the web, extract structured brand data, retrieve logos and styleguides, and monitor websites for changes. Built for AI agents.
Nuggetz turns messy workflows into bite-sized nuggets that progress from Manual to Assisted to Autonomous, handing work to the AI agents you already use.
A CLI package manager for open-source LLMs. Install, run and serve 10,000+ models from HuggingFace, Ollama and Mistral with a single command.
Dedalus Labs runs persistent Linux VMs with VM-level isolation that boot in 50ms, scale to zero, and bill per second. Purpose-built cloud infrastructure for AI agents.
Understudy is an open-source local AI agent for macOS that automates GUI apps, browsers, and messaging through natural language or teach-by-demonstration. Bring your own model.
Screenpipe captures your screen and audio 24/7, transcribes everything, and makes it searchable. It runs locally on macOS, Windows, and Linux, with optional team workflows.
Magnitude is an open-source AI agent with local models built in. Fully private, offline-capable, and works out of the box on any hardware. No API keys or token costs.
MIT-licensed AI agent server in one Rust binary. Sandboxed execution, hybrid vector + full-text memory, MCP tools, and Pi-inspired self-extending skills at runtime.
Prized lets ops, support, and finance teams build AI-powered internal tools on real company data without writing code. Guardrails, scoped access, and audit trail built in.
Polymorph runs thousands of AI-generated experiments per customer, automating push, email, and user segmentation to maximize conversion.
Apache Burr is an open-source Python framework for building stateful AI applications like chatbots and agents, with a built-in telemetry UI and Apache 2.0 license.
AI agent for planning company retreats — pick your city and dates, get venue recommendations, activity ideas, hotel options, and a full itinerary, all from a single prompt.
FutureSearch is an AI forecasting platform that scores above the superforecaster median on ForecastBench and ranks #1 of 226 on Metaculus FutureEval. Available via MCP server, Python SDK, or web app.
A full-featured local AI app with document chat, AI agents, multi-user support, and MCP compatibility. Works with any LLM, no setup required.
EdotEnv builds multi-step reinforcement learning environments at scale from financial market data, designed for agents that need to continuously hillclimb on harder research tasks.
Giskard v3 is a Python library for testing AI agents via scenario evals and automated vulnerability scanning. Covers prompt injection, groundedness, and multi-turn traces.
Floneum is a graph editor for building AI workflows with local LLMs and sandboxed WASM plugins. Write plugins in Rust, C, Java, or Go.
Publish authoritative domain identity as open JSON. The AI Domain Data Standard gives AI systems and automated agents a canonical, self-hosted source of truth about any domain — no SaaS required.
Juggler is an open-source GUI coding agent that visualizes tool calls, approval prompts, and branching conversation threads in a Miller column layout. Sessions persist across restarts, surviving even
AgenticSeek is an open-source, self-hosted AI agent that browses the web, writes code, and plans tasks — entirely offline. No API bills, no cloud dependency, just Docker plus a local LLM.
Stoa is an institutional marketplace for buying and selling GPUs and AI hardware with verified counterparties and real-time price discovery. Y Combinator S26.
Pretzel adds inline AI completions, a chat sidebar, and one-click model switching to Jupyter. Drop-in replacement with zero config.
Cloud-hosted AI agent that controls a remote desktop like a human, ships a /v1/predict REST API with 1,000+ OAuth app integrations, and scores 85.60% on OSWorld benchmark.
Discovered Materials builds AI agents that discover new materials for semiconductor chips, backed by a $9M seed round led by Lightspeed with YC P26 backing.
Squadbase connects your Git repo and deploys private, authenticated Streamlit or Next.js AI apps with RBAC, logs, and analytics in about three minutes. Free up to 5 apps.
A native macOS app that puts multiple AI companions in your conversations — they build live interactive dashboards and act on your machine through one unified API.
Aethene gives AI agents persistent, searchable memory. Store content, extract atomic memories, search semantically, and recall context across conversations with full MIT licensing.
Coasty is a YC S26-backed computer-use API that controls real desktops like a human. $0.05/v1/predict, 1000+ app integrations, 85.60% OSWorld benchmark.
One API for 300+ AI models. Opper AI is an EU-hosted, GDPR-compliant gateway with drop-in OpenAI SDK support and a built-in agent control plane.
See if ChatGPT, Claude, Gemini, and Perplexity recommend your brand. Enter any domain, get a free AI visibility score across major AI assistants.
A Python framework that simplifies building and managing multi-step LLM applications. Self-hosted, open source, with version control for prompts and built-in FastAPI.
Roxlit is an open-source launcher that connects Claude Code, Cursor, and Windsurf to Roblox Studio via Rojo and an MCP server, with Luau context packs for accurate game code generation.
Salestrics is an open MCP server and unified CRM that gives AI agents live access to pipeline, mail, Stripe, Slack, and 13 apps through 163 tools.
Pylar sits between AI agents and databases like Snowflake, Postgres, and BigQuery, letting you define data access policies and expose structured MCP tools without writing API code.
Connect your bank, cards, and investments to an MCP server. Ask Claude or ChatGPT about spending, net worth, and subscriptions.
Describe the infrastructure you want in plain text. DevOpsAgents provisions and manages it across AWS, Azure, GCP, and DigitalOcean using Terraform and Ansible.
Nuanced is a macOS desktop app that starts with your rough idea, clarifies it into a living spec, then executes through Codex while keeping you oriented as code evolves.
Chat with an AI to sort, search, and organize files across Google Drive, Dropbox, and OneDrive. Dynbox uses semantic vector search to find files by meaning, not just filename.
A free, local-first AI code editor built in Rust. Under 35MB, preserves your VS Code/Cursor themes and settings, and verifies code changes automatically.
One API to scrape the web into clean Markdown, crawl entire sites, and extract structured brand data. YC S26.
AI-powered investment research platform that surfaces stock opportunities using natural language queries and real-time market data.
A desk-size hardware console that runs local LLMs via llama.cpp and exposes an OpenAI-compatible API. No cloud, no latency, ships tuned.
Ainisa lets businesses build AI agents for WhatsApp, Instagram, and TikTok without code. Connect GPT-4 or Claude, automate customer support, leads, and bookings. Here's how it works.
Deploy MCP servers from GitHub to HTTP endpoints in 30 seconds. Open source AGPL-3.0, works with n8n, Dify, Voiceflow, and Langflow.
Browser extension that saves, bookmarks, and restores AI chat contexts across ChatGPT, Claude, and Gemini. Free plan available, no account required.
Exa is a search API built for AI agents. It delivers live web data, deep research, and structured content in a single endpoint — with claimed 20x better recall than traditional search.
HyperProbe is a YC S26-backed AI on-call agent that probes production services on incident, captures request context, and delivers it to Cursor, Claude Code, Codex, or opencode.
Hoplite deploys cloud coding agents that read your repo, run tests, drive browsers, and open GitHub pull requests. Built on Y Combinator S26.
cMCP is an open-source gateway that enforces MCP tool-call policy inside a Trusted Execution Environment, with signed TRACE claims for auditability.
Smooth CLI gives AI coding agents a goal-oriented browser that navigates the web via natural language instead of DOM clicks, cutting cost and latency dramatically.
Memcode is an agentic coding CLI that builds a persistent model of your codebase so the agent knows what came before, every session.
Daytona provides secure, sub-90ms cloud sandboxes purpose-built for running AI agents and AI-generated code at scale. Supports Python, TypeScript, Go, Ruby, and Java.
Hiver is an open-source runtime for AI agents that runs them in fast, isolated sandboxes with a full replayable timeline of every action — browser, file, network, and tool call.
VibePod is a zero-config CLI that runs Claude, Codex, Devstral, Gemini and other AI coding agents inside isolated Docker or Podman containers. No setup required.
Upstreet is an open-source framework for building and deploying AI agents using React components. The usdk npm package (v0.0.110) provides a TypeScript-first SDK with pnpm installation.
Open-source TypeScript framework for building production AI agents with built-in memory, MCP support, guardrails, RAG, voice, and a full observability console.
Relvy automates on-call runbooks for engineering teams. An AI agent with specialized telemetry tools analyzes logs, traces, and time-series data to help debug production issues faster.
open-context imports your ChatGPT history, analyzes your communication style via local Ollama AI, and gives Claude persistent memory through an MCP server. Open-source, MIT licensed.
Add 9 automation skills to Claude Code, Cursor, and 14 other AI tools. Reddit search, domain hunting, SEO optimization, and banner creation — in one install.
The drop-in trust layer for MCP — authentication, fine-grained authorization, consent, and audit for AI agents. Zero standing permissions. Enterprise-grade security.
Even combines a native terminal, real browser, and local LLM inference in one desktop window where you and your AI agents work side by side.
A native Mac app where you and multiple AI companions share live surfaces. Companions build interactive dashboards, tools, and apps mid-conversation that persist in your shared space.
Save, restore, and organize AI chat context across ChatGPT, Claude, and Gemini without sending data to third-party servers.
Athena gives any LLM persistent memory, structured reasoning, and cross-provider governance — so your context survives model switches and updates.
agent-data is a CLI that gives AI agents real-time, structured data via a unified API, replacing slow token-burning browser automation with fast targeted calls.
Memcode is a terminal-native coding agent that builds a persistent model of your codebase so every session starts with real context, not a cold start.
Tokenless is a YC S26 model router that fans out API calls to multiple models, selects the best responder, and cancels the rest — dropping costs by 50 percent with no quality loss.
A Model Context Protocol server that gives Claude Desktop access to Microsoft 365 mail, calendar, OneDrive and SharePoint via the Graph API.
PrivateClaw runs AI agents inside AMD SEV-SNP Trusted Execution Environments, keeping prompts and outputs encrypted at the hardware level—no more trusting hosted platforms with plaintext.
Stele is a hosted knowledge graph for AI coding agents. It tracks what your project has decided, tried, and learned—then surfaces it to every agent session before they act.
Drop a GitHub URL and get system architecture overviews, per-file summaries, and flow diagrams — powered by Gemini 2.5 Pro. Great for onboarding and codebase exploration.
A native macOS app that gives AI agents persistent semantic memory, so what you teach one tool is remembered by all others automatically.
A free, open-source AI agent that runs locally on Windows and macOS. Controls your PC by voice, automates Office docs, browses the web, and manages files.
Deploy full-stack apps, databases, and APIs from AI agents via CLI, Skills, or MCP. No Dockerfiles, no CI pipelines—just code.
Turn any idea or document into a professional launch video in seconds. Ozor.ai handles the editing — you just prompt. Free plan available.
UpTrain is an Apache 2.0 open-source platform that runs 20+ pre-built LLM evaluation checks, performs root cause analysis on failures, and self-hosts entirely on your infrastructure.
Lucid is a free web tool that lets you watch an AI model's internal concept activation layer by layer before it commits to an answer, based on Anthropic's Jacobian Lens research.
Line0 is the first AI pair programmer purpose-built for backend developers. It helps you build Express.js services with AI assistance, integrated directly into your workflow.
PlanBridge intercepts your AI coding agent's plan before any code is written, letting you annotate and approve line-by-line. MIT licensed, runs locally.
A privacy-first, open-source dictation app for Linux. Runs locally with whisper.cpp, VOSK, or OpenAI Whisper. GPU-accelerated via Vulkan, works on X11 and Wayland.
AgentPort lets you connect AI agents to Stripe, GitHub, Gmail and 50+ services with approval policies, full audit logs, and zero credential exposure.
Smooth CLI is a serverless browser agent API that lets AI agents like Claude Code navigate the web using natural language instructions, scoring 92% on WebVoyager at 7x lower cost than Browser Use.
Intercept, approve, and audit every LLM tool call with one Python decorator. Fail-secure by default.
Misatay is a VS Code extension that keeps you in charge while AI agents write code. Plan tasks, review changes, and merge with confidence.
Summon AI in any macOS app with one shortcut. Inset AI brings ChatGPT-class assistance to Photoshop, Terminal, Slack, and every other application on your Mac.
Tork intercepts every AI agent call, strips PII, enforces policies, and issues tamper-evident compliance receipts. 136 framework adapters, open-source SDKs, and a managed MCP governance server.
Wolffish is a local-first AI desktop agent that runs entirely on your machine with no cloud dependency. Built with a 15-module brain architecture, markdown-driven behavior, and full system access on
Max connects to data sources and syncs them into portable local storage, so AI agents query locally without API rate limits, latency, or per-query costs.
The RamaLama SDK brings containerized LLM inference to Python, letting you run models locally on any hardware that supports Docker or Podman.
Expose your entire Zabbix infrastructure — hosts, triggers, templates, problems — as 237 MCP tools. Works with Claude, Codex, Cursor, any MCP client. TOML config, OAuth 2.1 tokens, admin portal.
Graphiti is an open-source framework for building temporal context graphs that track how facts evolve over time, giving AI agents structured memory instead of flat chat history.
Open-source Next.js + Remotion template that lets AI coding agents churn out polished product launch videos in hours instead of days.
Traceforce monitors and controls AI apps, MCPs, and CLI agents across your device fleet. Open-source MCP X-Ray scanner generates SARIF reports for CI/CD pipelines.
Open-source memory graph for AI agents that indexes GitHub, Linear, Slack, and conversations into a temporal knowledge graph. 88.24% SOTA on LoCoMo benchmark, self-hosted, AGPL 3.0.
Local-first macOS app that records, transcribes on-device, and turns meetings into todos, topics, and people pages.
Build reliable background jobs and AI agent workflows without extra infrastructure. Automatic retries, flow control, and step-level observability.
Bike4Mind is an open-core AI workspace combining notebooks, autonomous agents, and a RAG knowledge engine across any model—OpenAI, Anthropic, Google, xAI, or local Ollama/vLLM.
Xorq turns ephemeral pandas scripts and ML pipelines into durable, composable, executable catalog entries any AI agent can discover and reuse.
Moltis is a self-hosted AI agent server written in Rust. One binary, sandboxed tool execution, multi-provider LLMs, voice, memory, and integrations with Telegram, Discord, and MCP tools.
Prized lets ops, support, and finance teams build and ship secure internal tools with AI. No engineers needed—connectors are pre-scoped with audit trails.
Smithery.ai is a marketplace of MCP servers — connect AI agents to hundreds of tools and services. Auth, credentials, and sessions handled for you. Pay-as-you-go pricing.
Colibri is a pure-C MoE runtime that streams 744B-parameter GLM-5.2 from disk to consumer hardware. Zero Python, zero GPU required — ~25 GB RAM and you are running a frontier model locally.
mcp-use is the full-stack MCP framework that lets you build MCP Apps for ChatGPT and Claude, plus MCP Servers for AI agents, all in TypeScript or Python.
osa-mcp is an MCP server that auto-discovers all scriptable macOS apps and generates ~700 tools with real descriptions. Connect to Claude Desktop or any MCP-compatible AI in seconds. BSD 3-Clause.
Enterprise-grade local-first memory for AI agents. Runs on your infrastructure, supports MCP and CLI, GDPR and EU AI Act compliant.
Drop a YouTube URL or upload a video and get a full blog post, article, or social media copy. Built with AI, one-time pricing.
A user-owned data wallet giving AI agents rich persona, memory, and preferences across 1000+ OAuth sources — privacy-first, SDK-ready.
Open-source browser automation that uses vision LLMs to navigate websites, fill forms, and complete workflows without brittle XPath selectors.
Persistent memory across Claude Code, Cursor, and ChatGPT. Mengram stores facts, episodes, and procedures that survive every /clear.
A scheduling API that handles timezone math, DST edge cases, and conflict-safe bookings — so you keep your UI and delegate the hard parts to Kalendis.
Vett is an open-source CLI and hosted platform that scans AI agent skills for malicious code, signs verified ones, and runs each agent in an isolated Firecracker microVM.
Inconvo is an open-source YC S23 platform for building chat-with-data agents on production databases — without SQL. Safe queries, row-level permissions, and structured outputs.
Open-source MCP server that scans Python AI agents for EU AI Act compliance, generates code fixes, and adds cryptographic audit trails via AirTrust.
Browserable is an open-source library that gives AI agents reliable browser control — navigation, form fills, clicks, and data extraction — scoring 90.4% on the Web Voyager benchmark.
Run one AI assistant across chat, tasks, teams, and channels — VeryBot is a self-hosted work platform you install via npm, with MCP support and no data leaving your server.
Phi-3-MLX brings Microsoft's Phi-3.5-vision and Phi-3.5-mini models to Apple Silicon Macs via MLX. Install with pip and run locally in minutes.
A native policy enforcement layer for AI coding agents that intercepts actions and evaluates them against OPA/Rego policies, blocking unsafe operations before they execute.
GlitchWard LLM Shield detects and blocks prompt injection attacks before they reach your AI model. Sub-50ms response with a 6-layer analysis pipeline.
Waffy is a free, open-source Chrome extension that automates web tasks using AI. Tell it what to do in plain English and it clicks, types, and scrolls through workflows for you. No backend, no
Vexp builds a dependency graph from your AST and serves only the relevant code subgraph to AI agents, cutting token usage by 65-70% with zero network calls.
Autonomous AI agent that reads Jira or ClickUp tickets, analyzes your codebase, and opens production-ready pull requests on GitHub. Docker one-liner setup.
One Rust binary that runs a persistent personal AI agent on your own hardware. Sandboxed execution, voice, memory, 30+ integrations, no npm or Node.js required.
Open-source Rust-powered SQL query and AI compute engine for data-grounded applications and AI agents. Federated queries, acceleration, RAG, LLM inference.
uLog.ai uses ChatGPT to guide reflective journaling sessions, asking adaptive questions and building searchable personal timelines from your answers.
MCP Vault (mcpv) is a Python-based performance booster for AI agents. It forces GPU acceleration, caches tool responses, and deduplicates repeated requests to cut token costs by 90%.
A browser-based studio for building MCP agent systems. Design multi-agent workflows visually, run them in Pyodide WASM, then export as production Python servers.
FizzBee is a free web app that uses formal verification to elicit requirements, uncover gaps, and generate precise specs for AI coding agents — fewer iterations, more correct code.
LayerNext deploys AI agents that automate AP, bank reconciliation, bookkeeping, and CFO reporting across QuickBooks, NetSuite, SAP, and ERPs. Plans from $79/month.
Drop an OpenAPI spec or Postman collection and get a working AI agent that calls your APIs using natural language.
Screenpipe runs locally on your machine and records your screen 24/7, transcribes audio, and gives AI agents a searchable memory of your work. Install with npx, query with natural language.
Point your coding agent at your repo and docs; it opens a docs branch, your team reviews on the rendered page, publishing is a commit. Runs on Cloudflare Workers with no platform fee.
Run autonomous penetration tests with local LLMs. LLMtary chains recon, exploitation, and reporting—Ollama, Claude, GPT-4, and Gemini supported.
An MCP server that brings async-first work practices into your editor — decision docs, status scoring, async standups, and role guidance for distributed teams.
InsForge is an open-source cloud platform built for AI coding agents. It provides a model gateway, Postgres, auth, storage, edge functions, and hosting — all operable end-to-end through CLI.
DataFrey connects Snowflake to Claude Code via MCP, letting you query databases with natural language using a built-in text-to-SQL planning agent.
Secure credential injection, policy enforcement, and audit trails for AI coding agents. Open-source Go CLI with Homebrew install, MIT licensed.
SwarmZero is an open-source Python SDK for building, orchestrating, and monetizing AI agent swarms. Supports OpenAI, Anthropic, Mistral, Gemini, Ollama, and more.
SourceWizard is an open-source CLI that uses AI to install and configure SDKs automatically — handling middleware, env vars, and package setup so you stop reading READMEs.
VibeBrowser MCP connects Claude Code, Cursor, VS Code Copilot, and OpenCode to your actual Chrome profile — cookies, sessions, and all — via a hosted relay. No local browser install needed.
Coasty ranks #1 on OSWorld (85.60%) and ships a production computer-use API with 1,000+ OAuth integrations, MCP server, and a $99/mo flat Unlimited plan.
Superfire packages any MCP-powered tool into a shareable link so visitors can try the agent instantly, no install required.
Voker is a YC S24-backed analytics platform that gives AI agent builders full visibility into sessions, corrections, and resolution rates via a drop-in SDK.
Skyvern is an open-source AI agent that automates browser workflows using vision LLMs and Playwright, replacing brittle XPath selectors with adaptive visual navigation.
OpenGAP defines AI agents as a folder of YAML and Markdown files. Clone a repo, run opengap, and any AI framework becomes your runtime.
MIT-licensed visualization language from Microsoft that compiles simple chart specs into Vega-Lite, ECharts, and Chart.js, with an MCP server for AI agents.
Open-source edge AI engine that runs local LLMs on iOS, Android, and wearables using custom quantization, CPU/GPU kernels, and an OpenAI-compatible API.
Deploy Bun or Python functions as versioned, sandboxed HTTP endpoints with 800+ OAuth integrations. Velane gives AI agents the integration layer they actually control.
MarinaBox provides secure, isolated browser and desktop environments for AI agents. Docker-based, Python SDK included, works with Playwright and Selenium.
Structured AI design review by Claude, ChatGPT, Gemini, and Grok. Submit your code, get parallel critical analysis with confidence levels, and ship with confidence.
Setoku is a self-hosted MCP knowledge server that gives AI agents the metric definitions, schema context, and gotchas they need to query company data correctly — no AI runs on the server.
PycoClaw runs a full OpenClaw-class conversational AI agent on an ESP32-S3 microcontroller for under $5 in hardware. Features MicroPython, LVGL, hybrid memory, and OTA updates.
AG2B runs the full agent loop in the browser with tools, scopes, and typed Zod handlers — no backend required for agent logic.
Traycer connects Claude Code, Codex, Cursor and other coding agents in one shared workspace with unified context, agent-to-agent loops and real-time collaboration.
Smooth CLI replaces low-level browser automation with natural language commands, letting AI agents handle web tasks without context-window bloat or token waste.
CyberCage is an enterprise AI security platform that provides MCP server discovery, approval workflows, and tool usage tracking for teams building with AI agents.
Pick up an AI coding session from Claude, Gemini, Codex, or 13 other tools and hand it off seamlessly — no context lost.
A complete STT + LLM + TTS + VLM pipeline running on Apple Silicon with sub-200ms latency. 40 macOS actions via voice, local RAG, no cloud, no API keys.
How EGC connects 20+ AI coding tools—Cursor, Claude Code, Copilot, Aider and more—with shared persistent memory, encrypted locally.
Open source AI automation platform with 280+ MCP tools, self-hosted deployment, and a visual flow builder — alternative to Zapier with AI-native features.
Open-source web scraper that outputs clean LLM-ready Markdown with structured JSON, MCP server, and Docker support. 74K GitHub stars.
Leap is an AI developer agent that builds and deploys full-stack apps with real backends to your own AWS or GCP cloud. Connect GitHub, review PR-style diffs, and deploy via Encore.ts.
Open-source tool that uses LLMs to extract JSON from PDFs, DOCX, and images. Custom schemas, local model support, and self-hosted.
Six Playwright tools let AI agents open browsers, read pages, and act. 55% lower cost than alternatives.
Argmin AI optimizes model selection, prompts, and routing to cut AI agent costs by up to 10x. Prompt compression, RAG tuning, and FrugalGPT-style routing.
Sphinx is a data-native AI copilot for Jupyter notebooks that understands data science workflows. It runs an agentic loop cell-by-cell, reads charts, and connects to Snowflake and Databricks.
Driftcop detects drift, injection attacks, and CVEs in MCP servers with SigStore tracking — the SAST tool your AI agents need.
YourMemory gives AI agents persistent, self-improving memory using biological Ebbinghaus decay curves, graph linking, and consolidation. +16pp better recall than Mem0 on LoCoMo.
ActivePieces turns its 280+ automation pieces into MCP servers your AI agents can use. Self-hosted, TypeScript-based, MIT licensed.
Minicor deploys AI agents that automate legacy Windows desktop systems like EHRs and ERPs, self-healing when UIs change. SOC 2 and HIPAA compliant.
Open-source AI code review tool that analyzes git diffs and generates structured reports with severity ratings. Supports Anthropic, AWS Bedrock, Ollama.
pg-mcp bridges PostgreSQL and AI agents through the Model Context Protocol, exposing schema introspection, read-only queries, and EXPLAIN analysis as tools any LLM can call.
MemoryStack gives AI agents persistent memory, knowledge graphs, and semantic search. Free tier includes 500 memories/month with Python and Node.js SDKs.
Port42 is a native macOS app where multiple AI companions and humans coexist in shared chat channels, building real-time visualizations and tools together. Open source, MIT licensed.
Arrakis is an open-source sandboxing service that runs AI agent code in secure MicroVMs with snapshot-and-restore, a REST API, Python SDK, and MCP server support.
MCPJungle is a self-hosted gateway that centralizes multiple MCP server connections behind a single endpoint, replacing scattered per-client configs.
Python tool that polls Zotero for new papers, syncs PDFs to reMarkable, extracts highlights, and spins up Claude Code research agents from your reading.
AgentLens is an open-source, self-hosted observability platform for AI agents. It records every tool call, lets you replay failures step-by-step, and uses AI to diagnose root causes.
OMS defines the .mg container format for immutable, content-addressed AI agent memory. Open standard with SHA-256 hashing, COSE signing, and three specs covering storage, querying, and output.
CADAM transforms natural language and images into 3D CAD models directly in the browser using OpenSCAD WASM. Free, open-source, no install required.
MCP Jetpack connects Cursor AI to GitHub, Linear, Atlassian, Notion and more without managing individual MCP servers. Auto-discovers tools, bypasses Cursor 50-tool limit.
Connect AI agents directly to UnDatasIO document parsing via an MCP server. Upload files, trigger parsing, and poll results through a stateful interface without writing workflow boilerplate.
TypeWhisper is an open-source, local-first speech-to-text app for macOS and Windows with app-aware workflows, automatic correction learning, and free core dictation under GPLv3.
Construct gives your AI agent a persistent work OS — memory, files, browser, terminal, schedules, workflows, and 1,000+ connected apps in one supervised cloud workspace.
NetFabric deploys autonomous AI agents that investigate, coordinate, and act across your entire network infrastructure. ETH Zurich spin-off with $2.2M pre-seed and Innosuisse grant.
CapaKit is a free runtime and CLI toolkit for building, packaging, and running AI app Kits — self-contained workloads with MCP tools, voice agents, and more.
Sonarly connects to Sentry, Datadog, and Slack to triage alerts, identify root causes, and trigger AI fixes — cutting a 180-alert daily firehose down to 5 actionable issues.
Give AI agents their own disposable microVMs with hardware isolation. Sub-second boot, browser sandbox, file mounting, Python SDK and CLI. Apache 2.0.
Run AI agents in isolated, resumable MicroVM sandboxes with persistent filesystems, live run events, and a CLI that works with Claude Code, Cursor, and Codex.
A free, ODbL-licensed nutrition database with 320K+ foods, LLM-powered web search, and a macro tracking iOS app.
A secure persistent AI agent server in Rust. One binary, sandboxed execution, multi-provider LLMs, voice, memory, and MCP tools. Runs locally.
Writtte is an AI-assisted editor that helps founders, bloggers and creators turn ideas into polished articles for WordPress, Medium, Substack and Markdown — all distraction-free.
Coasty gives developers a $0.05 per predict API to control a real browser or desktop, with 1,000+ native app integrations for Gmail, Slack, GitHub, and more.
BrowserOS is an open-source Chromium fork with a built-in AI agent. BrowserClaw gives your AI agents control of a real browser signed into your accounts. AGPL-3.0, 12K+ stars.
Generate code from natural language, test in a sandbox, then deploy as a REST endpoint in one click with Structured Blocks.
MinusX adds an AI side-chat to Metabase that answers any business question in plain English, using your existing dashboards and custom metrics as context.
Stagewise is an open-source agentic IDE that embeds an AI coding agent directly into your browser tab, with full access to console and debugger. Supports 30+ models.
Verdent AI is an agentic coding suite that runs multiple specialized agents in parallel, plans tasks with a built-in Plan Mode, and ships production code via VS Code or JetBrains.
A free browser tool showing which internal concepts a language model holds mid-computation, using Anthropic's Jacobian lens to let you observe reasoning layer by layer.
Cq is an open source framework from Mozilla.ai that lets AI coding agents share what they learn so others avoid repeating the same mistakes.
Bloomiro connects your website to Google Search Console and AI search results, then turns the data into prioritized SEO and GEO tasks you can run through an MCP agent in Cursor, Claude, or ChatGPT.
Matrix OS gives AI agents their own persistent hosted workspace with terminal, files, apps, and Postgres — accessible from any browser after your laptop closes.
RCLI is a local voice AI for macOS running a complete STT + LLM + TTS + VLM pipeline on Apple Silicon. Sub-200ms latency, 40 voice-controlled macOS actions, local RAG over your documents.
1Code is an open-source coding agent client that runs Claude Code and Codex with a visual UI, git worktree isolation, and background cloud sandboxes.
Run AI agents locally on macOS/Linux with a personal CFO, Mac health doctor, and DJ. Drop-in skills, P2P model sharing, MCP server.
mcp-use is the open-source MCP framework with TypeScript and Python SDKs. Build MCP servers and apps, test with the inspector, and deploy to Manufact Cloud.
Fluree DB is an open-source graph database that speaks SPARQL, JSON-LD, and Cypher natively, with built-in vector search and AI agent memory.
Stop AI code hallucinations. GitMCP creates an MCP server for any public GitHub repo, giving AI coding tools like Cursor and Claude accurate, up-to-date documentation and code.
Superset is a desktop app that orchestrates multiple AI coding agents in parallel using git worktrees. Built with Rust, it works with Claude Code, Codex, and other CLI agents.
Run any AI agent in a zero-latency kernel-enforced sandbox with zero setup. Nono isolates agents and their tools with least-privilege policies, credential proxying, and L7 API filtering.
Open-source MCP server providing persistent semantic memory for AI agents. Supports Postgres + pgvector, OAuth 2.0, and 20 tools for storage, retrieval, and knowledge graphs.
Fireproof is an embedded JavaScript database with encrypted sync, Git-like versioning, and React hooks for building offline-capable apps.
An AI content brain for creators that lets you chat with TikToks, YouTube videos, podcasts, and PDFs using a visual board interface. From $6/month.
AI-powered platform that generates SEO-optimized blog posts automatically for SaaS growth. Turn content creation from bottleneck into autopilot.
Free, independent hub scoring B2B data and AI-agent APIs against verified ground truth. Covers lookalikes, enrichment, funding, TTS, and STT with public methodology, OpenAPI 3.1, and MCP access.
Runtime uses predefined DOM attributes as skill gateways to automate browsers, sidestepping expensive LLM reasoning at execution time.
Build and run agentic workflows across channels with YAML, MCP tools, memory and RAG. Step-by-step setup using the Hexabot CLI.
QueryVeil runs entirely in your browser. Drop in a CSV and the AI investigates your data without ever uploading it — DuckDB handles processing locally.
Port42 is a native macOS app where multiple AI companions collaborate with humans in the same conversation, build real-time ports, and run CLI agents inline. Open source and end-to-end encrypted.
Cognee gives AI agents persistent long-term memory using a knowledge graph. Self-hosted, Apache-2.0, and works with Python 3.10+.
WonderTwin gives your AI agents local, behavioral clones of third-party APIs — Stripe, Twilio, Resend, and more — that run offline and mirror production failures. Open-source MIT.
Freestyle gives AI coding agents their own Linux VMs with git-backed persistence, sub-second provisioning, and GitHub sync. Open-source option available.
Coasty is a computer-use AI agent ranking 85.60% on OSWorld, with a REST API at $0.05 per predict, MCP server, and 1,000+ OAuth integrations for $99/month flat.
papercrane-cli gives Claude Code authenticated access to your data sources and turns business questions into live, shareable dashboards — no coding required.
MLJAR Studio is a desktop app where you chat with your data in natural language and get reproducible Python notebooks. AutoML, pandas, and database connectors included.
Membrane gives AI agents typed, revisable memory instead of append-only transcripts or flat vector stores. Five memory layers, trust-aware retrieval, and Go/TypeScript/Python SDKs.
Open-source tool that turns long podcast episodes into short clips with face tracking, burned-in captions, and 26 MCP tools for agent-driven workflows.
MirrorNeuron is an open-source runtime that keeps long-running AI agents recoverable, resumable, and locally private — without Airflow, Temporal, or cloud infrastructure.
Find the best MCP servers for Claude and AI workflows. MCP Review is a free, open-source directory with community-driven ratings and reviews, MIT licensed.
AgentWatch is an open-source proxy firewall that sits between AI agents and LLM providers, blocking requests that would blow past your budget. Two lines of code, 10 providers, sub-10ms latency.
Cybara is an open-source, self-hosted AI agent OS built on Bun with browser automation, MCP support, multi-channel messaging adapters, and encrypted wallet controls.
OpenTrace connects Claude Code and Cursor directly to your production observability data via MCP. Self-hosted, no Elasticsearch, runs on a $4/month VPS.
Set up DenchClaw, an OpenClaw-based local AI CRM that runs on your Mac. Manage leads, automate outreach, and chat with your data - no cloud required.
Qurt is a free, open-source desktop AI coworker for macOS and Windows that researches the web, writes drafts, runs commands, and reads your files. BYOK with no subscription.
Shuru boots lightweight ephemeral Linux VMs on macOS using Apple Virtualization.framework, giving AI agents a disposable sandbox to run code safely without touching the host.
Julie is an open-source desktop AI that sees your screen, responds by voice or text, and autonomously controls your browser or computer via Puppeteer and JXA.
Coasty is an AI agent API for computer-use automation. Score 85.60% on OSWorld benchmark, 1000+ app integrations, and $0.05 per predict call. YC S26.
SUTRA.team runs 15 AI specialists as a deliberating council, using the Sammā Suit 8-layer security framework to synthesize decisions for autonomous agent teams.
Connect, secure, control and observe every MCP connection, AI skill and agent session. Speakeasy is the control plane for enterprise AI governance at runtime.
Free tool to audit your brand's AI search visibility across ChatGPT, Gemini, Claude, Perplexity, Grok, and Google AI with actionable GEO reports.
Compare response time, throughput, and cost across OpenAI, Claude, Gemini, and any OpenAI-compatible local model with this browser-based benchmark tool.
Build real-time voice and multimodal AI agents with Pipecat, the open-source Python framework from Daily.co. Supports 50+ STT, LLM, and TTS services.
An MCP server that gives AI agents full control of a real iPhone through macOS iPhone Mirroring. Screenshots, taps, swipes, and app control.
Run GGUF models on your own hardware with a native .NET inference engine. CLI, web UI, and Ollama/OpenAI-compatible APIs on Windows, macOS, Linux.
Bodhi App is a free, open-source desktop app that runs open-source LLMs locally on macOS, Windows, or Linux with an OpenAI-compatible API and built-in Chat UI.
Captain connects to S3, SharePoint, and 20+ sources to index video, images, PDFs, and audio, then serves AI agents via a clean REST API or MCP server. YC W26.
GitHits is an open-source CLI and MCP server that gives AI coding agents version-aware indexed access to public open-source code, package docs, and changelogs.
Cardboard is a browser-based AI video editor that cuts, captions, reframes and composes footage from a text prompt. From Y Combinator W26.
ScopeGate is an open-source MCP proxy that gives each AI agent exactly the access it needs — nothing more. Connect Google services, define granular OAuth scopes, and get an MCP endpoint URL in
ContextVault stores shared memories once and makes them reusable across your team. A hybrid retrieval layer for AI that survives sessions.
Control Claude Code and Codex by voice from your phone, car, or watch. Self-host free with Java 21 or use the managed cloud.
SysModeler uses AI to generate and understand SysML diagrams without manual effort or tool complexity. A practical look at AI-powered MBSE.
Frontier coding agent powered by open models like GLM and DeepSeek. Runs in your VPC, code never leaves your infrastructure. 75.5% Terminal-Bench 2.1, 3x cheaper than Claude Code.
Deploy a self-hosted RAG pipeline with built-in MCP tools for PostgreSQL, MSSQL, MySQL, SSH, and filesystem indexing. Works with Claude Desktop, Cursor, and any OpenAI-compatible client.
Tambo is an open-source React toolkit that lets AI agents render dynamic UI components — charts, forms, maps — streamed as props in real time. 11K GitHub stars, MIT licensed.
A full-featured desktop AI assistant that runs locally, supports multi-user workspaces, and connects to any LLM provider. MIT licensed, with built-in agents and document chat.
A free web tool that visualizes internal activation patterns of language models layer by layer as they process a prompt, letting you observe and manipulate concepts mid-computation.
Skillscript is a Makefile-inspired DSL that lets AI agents author reusable, human-approved workflows with built-in security guardrails and MCP connectivity.
Statey is an MCP-native shared database that lets Claude, Cursor, and ChatGPT share structured records across every conversation. Hosted, free to start.
Cordon is an open-source security gateway that sits between AI agents and MCP servers, enforcing policy-based access control with human-in-the-loop approvals.
Transform any codebase into a single navigable Markdown doc for AI tools. Go CLI, MIT licensed, with file filtering and clipboard support.
Engram gives AI agents persistent, searchable memory with a 5-line Python SDK. SQLite-backed, self-hosted, zero-config, MIT licensed.
Stoneforge is an open-source TypeScript platform for orchestrating multiple AI coding agents in parallel. It features a web dashboard, git worktree isolation, and automated merge review.
WikiTimeline converts any Wikipedia article into a beautiful interactive timeline using Google Gemini AI — compare historical figures, events, and discover connections between them.
An MCP server that gives AI agents real-time football scores, player lookups, and match summaries from Vublox. No API key needed.
Semble indexes any codebase in ~250ms and answers natural-language queries in ~1.5ms on CPU, using 98% fewer tokens than grep-plus-read. MIT-licensed, no API keys needed.
A browser-based tool to connect to multiple Model Context Protocol servers, browse tools and resources, and chat with AI models.
Connect any database to AI agents via auto-generated MCP or REST APIs. Supports PostgreSQL, MySQL, ClickHouse, Snowflake, BigQuery, and more with built-in PII protection.
Aegize is a runtime governance layer that sits between autonomous AI agents and the tools they control — adding identity, policy enforcement, permissions, and audit logging.
A lightweight observability tool that wraps your existing LLM calls and surfaces traces, token costs, latency, and errors in a single dashboard.
Open-source platform for securing AI agents with fine-grained policies, runtime enforcement, and tamper-evident audit logs. Apache-2.0.
EchoStack evaluates conversations, forms, emails, and tickets — returning qualification status, extracted fields, and next action — or executing the outcome automatically via playbooks.
OpenBrowser is an open-source AI browser automation framework using a CodeAgent architecture to control any browser via CDP, with MCP and Python 3.12+.
Nimble is an MCP server that lazy loads tools on demand instead of flooding your context window with every tool schema. Over 99% token savings on initial MCP load.
A local document memory server for AI agents. Store notes, code snippets, and task context in LevelDB with a fast CLI and optional daemon.
AgentGram is an open-source, self-hostable social network built for AI agents. Features Ed25519 cryptographic auth, semantic search, Python/TypeScript SDKs, MCP server, MIT license.
HNSignals transforms HN comment threads into executive-level intelligence briefs with structured Hook, Gist, Debate, and Verdict sections.
Open-source app that runs any ComfyUI workflow with zero setup. Auto-installs custom nodes, missing models, and isolates each workflow in its own environment.
A local-first desktop AI workspace that keeps data in SQLite, supports cloud and local models, and organizes long-running AI workflows with branching sessions, knowledge bases, and extensible applets.
Onit SideKick docks an AI chat sidebar to any Mac app, letting you highlight text or load a window as context for Claude, GPT, Gemini, or local Ollama models.
AMP is an open-source memory server for AI agents with MCP native support, SQLite storage, D3.js visualization, and a 3-layer brain architecture.
Papr gives AI agents persistent memory, knowledge connections, and context intelligence. Ranked #1 on Stanford's STaRK benchmark at 91% accuracy.
An open-source MCP server that gives AI agents 18 tools for querying FIFA World Cup 2026 data – matches, teams, venues, standings, and more – with no API keys required.
ReadyBase transforms AI-generated outputs, raw data, and research into polished PDF and PNG documents without templates. Free to start.
Port42 is a native macOS app (Apple Silicon, macOS 14+) where humans and AI companions coexist in the same conversation, build tools together in real time, and share an end-to-end encrypted open
Linkly AI indexes your local documents and exposes full-text search via an MCP server and CLI, letting any AI coding agent query your knowledge base directly.
Replace 'hub' with 'ingest' in any GitHub URL to instantly get a codebase digest optimized for LLMs. No sign-up, free, open-source.
Dedalus Labs launches persistent Linux machines that boot in under 250ms with full VM isolation. Connect any LLM to any MCP tool without YAML or Dockerfile config.
Webpages average 50K+ tokens. BotBrowser strips the bloat and returns clean markdown, saving 90–95% of tokens for LLM agents. MIT licensed.
Tabstack gives AI agents structured web data and browser automation via a single API. Built by Mozilla, it handles the messy crawling layer so you can focus on the agent logic.
Cua (YC X25) packages the sandbox, SDK, and benchmarks you need to build, evaluate, and deploy AI agents that control full desktops — macOS, Linux, Windows.
WebMCP Core (@keak/webmcp-core) crawls any URL, extracts forms and API calls, and auto-generates structured tool definitions compatible with Chrome 146 WebMCP. MIT licensed.
Ragtoolina adds semantic codebase RAG to any MCP-compatible AI coding tool, cutting token usage by 63% with one line of config. Self-hosted option included.
Superset orchestrates multiple AI coding agents in parallel using git worktrees, letting you monitor and switch between tasks without context-switching overhead.
Free, open-source voice dictation for macOS. Whisper + local LLM polish that adapts tone per app—Gmail gets formal, Slack stays professional, LINE is casual.
Turn one-sentence prompts into production-ready AI function endpoints. Weco AI Functions generates structured, scalable REST APIs with web-search and multimodal support.
Local-first AI memory tool with 74.8% zero-LLM benchmark on LoCoMo. No API keys, no cloud, AGPL-3.0 — runs fully offline.
TaskPeace is a ranked task queue that AI coding agents like Claude Code and Cursor pull from over MCP. Free tier + $10/mo Pro.
Shellular gives you a full agentic dev environment on your phone — Claude Code, Codex, Cursor, and OpenCode with a real UI, terminal, file browser, and browser DevTools.
A visual IDE for building complex AI agents and prompt-chaining workflows. Ships with a desktop app, TypeScript library, and integrations for GPT-4, Claude, and more. MIT licensed.
Vessel is an open-source browser built from the ground up for AI agents. It provides durable sessions, MCP control, and BYOK for autonomous web browsing on Linux, Mac, and Windows.
Qualcomm's open-source GenAI runtime for Snapdragon. Runs GGUF models from Hugging Face locally on Hexagon NPU, Adreno GPU, or CPU via CLI, Python, or OpenAI-compatible server.
Polos is an open-source runtime for AI agents that handles sandboxed execution, durable workflows, human-in-the-loop approvals, and built-in triggers. Write agents in Python or TypeScript and ship to
owthorize is a synchronous Node.js guard that parses SQL, HTTP, shell, and filesystem tool calls into typed shapes and evaluates them against rules — catching destructive AI agent actions before they
Beitha is an AI-powered Chrome extension that opens tabs, fills forms, clicks buttons, and extracts data using natural language — no new browser required.
Open-source incremental data framework that keeps AI agents grounded in fresh, structured context. Only the delta is reprocessed — no stale batches, zero re-processing.
bVisor spins up an isolated bash sandbox in ~2ms using Linux syscall interception, no VMs or remote infra needed. Built in Zig, ships as a TypeScript SDK.
peerd is a browser extension that turns Chrome or Firefox into a full AI agent harness with sandboxed compute, WebRTC P2P sharing, and no backend.
Sipp is an open-source WebGPU runtime for GGUF AI models, built with Rust and C++. It runs models in-browser with zero dependencies and extends to native inference and cloud provider routing via a
Velda turns any dev environment command into a cloud batch job. No Dockerfiles, no image registries, no Kubernetes manifests — just prefix with vrun and scale.
A lightweight React SDK from Stytch that detects AI agents on your site and lets you serve different content to bots vs. humans. Easy drop-in components.
Intercept, audit, and validate AI agent tool calls before execution. Viberails adds a security layer between AI agents and their operations with under 50ms latency.
Commander is a Chrome extension that uses AI to automate web tasks via natural language — "go to YouTube, search for IPL, play the first video."
Validate AI agents without building your own test infrastructure. Automate coverage, catch regressions early, and verify vendor systems without code access.
Access H100, H200, B200 GPUs on demand from 30+ cloud providers via a single API. Compare prices, deploy clusters up to 64 GPUs, no long-term commitment.
Sipp is an open-source WebGPU runtime for running GGUF AI models directly in the browser with no server calls. Built with Rust and C++, it delivers 3x faster inference than comparable approaches with
Klavis AI provides MCP integration platforms that let AI agents use 100+ prebuilt tools reliably at any scale, with self-hosting, Python and TypeScript SDKs, and REST API. Apache-2.0 licensed.
INAYA forecasts commodity prices 24 months ahead, models BOM-linked costs, and produces audit-ready decision records for procurement teams.
OpenAgents Workspace is an open-source platform where AI coding agents share threads, files, and a live browser — letting you coordinate multi-agent workflows from a single URL.
A lean Python toolkit for building AI agents from simple YAML specs — supports OpenAI, Anthropic, DeepSeek, and more.
Vesper is an MCP-native tool that lets AI agents autonomously search, download, clean, and export datasets from research papers, web pages, code repos, and more.
Policy engine, EDR, and signed audit chain in one binary for securing AI agent tool calls, file access, and network egress.
Drop YouTube videos, PDFs, docs or audio and get production-ready SKILL.md files for Claude, Cursor, Copilot and 30+ AI agents.
Cloud-hosted AI agents that control a real browser to automate web workflows — prospecting, outreach, social media, and lead generation without code.
Davia auto-generates interactive wiki-style docs from your codebase via AI coding agents. Docs are editable, include visualizations, and sync to a collaborative workspace.
Serverless AI search platform by Chroma. Supports vector, full-text, regex, and metadata search on object storage. 28k GitHub stars, Apache 2.0, SOC 2 Type II.
Nuanced uses static analysis to generate enriched function call graphs for Python packages, giving AI agents structural code understanding.
Open-source framework for building durable AI workflows with human approval, structured outputs, and versioned execution. Works with Node.js and Go.
Paper Lantern MCP brings 2M+ research papers to AI coding agents. On-demand academic knowledge for Claude Code, Copilot, and Cursor.
Test and benchmark AI web agents against Mind2Web, WebArena, and AgentInstruct datasets with ground-truth trajectories and sandboxed websites. Built by a16z-infra.
FlowScript is an open-source TypeScript SDK and query engine for AI agent reasoning. 21 typed markers encode decisions, tradeoffs, and causal chains — D3 web editor, no network dependency.
Save X bookmarks with one click, organize with folders and tags, and query them from Claude Code via a built-in MCP server.
Ratschn is a native macOS dictation app that runs Whisper locally on Apple Silicon, transcribing voice to text entirely offline with no subscriptions.
Nelly is a cross-platform AI agent builder that lets you create, test, and deploy teams of agents using natural language — no coding required. Currently in beta for macOS, Windows, and Linux.
Outlit unifies customer data from your entire stack into a single agent-ready profile. YC-backed, $500/mo Team tier, SOC 2 Type II certified.
RapidFire AI is an open-source experiment execution framework for LLM customization — RAG, fine-tuning, and post-training — delivering 16–24x throughput over sequential pipelines.
Kita uses vision AI to parse bank statements, tax returns, and P&L documents for credit underwriting, handling messy photos and scans that break standard OCR tools.
Cardboard is a YC W26 AI video editor that runs entirely in your browser. Describe an edit in plain text and it cuts, captions, reframes, and composes your footage.
Build and run custom internal tools in minutes with AI agents that understand your business, workflows, and operations. Free plan available.
Lightbox records every tool action an AI agent takes, immutably and locally. Debug, replay, and audit autonomous agents with forensic precision.
AgentLair gives AI agents a persistent email identity, encrypted credential storage, and namespace isolation via a single API — no OAuth, no human in the loop.
TesterArmy is a YC P26-backed agentic QA platform. Describe test flows in plain English and let AI agents run end-to-end checks on web and mobile, with Slack alerts and CI integration.
Kampala from Zatanna captures the real network traffic behind any web portal and exposes it as a stable, callable API endpoint for AI agents.
Open-source AI agent monitoring platform. Full observability into what is failing in production, issue detection, semantic search, and eval alignment tracking.
Enact is an open-source registry and runtime for discovering, verifying, and running AI-executable tools with cryptographic signatures and policy enforcement.
TengineAI provides a remote MCP server that exposes your HTTP APIs as tools to AI runtimes, handling auth, execution, and observability.
Forge is a Python reliability layer that wraps self-hosted LLM backends with tool-calling guardrails, pushing an 8B model from 53% to 84%+ success on agentic tasks.
Open-source MCP server that scores AI agent output quality, catches safety failures like PII leaks, and enforces cost budgets across traces.
A lightweight self-hosted tool that runs multiple llama.cpp server instances simultaneously, with OpenAI v1 API compatibility and a real-time GPU dashboard.
Local-only AI memory that scores 74.8% on LoCoMo benchmarks with zero cloud dependency. AGPL-3.0, supports Claude, Cursor, 16+ agents.
Build and deploy AI agent workflows across chat, voice, and API channels with this open-source Node.js framework featuring YAML workflows, MCP support, and RAG.
Generate MCP-compatible tools from any OpenAPI spec and connect to Claude, Cursor, or any MCP client. Official Taskade MCP server with 62 built-in tools.
A Chrome Extension that exposes your browser as a remote MCP server, letting any AI agent drive web tasks without context switching. BYO-subscription model.
Voygr gives AI agents and apps fresh, queryable place profiles combining accurate location data with news, events, and web context that Google Maps APIs cannot.
Connect AI agents to the web through browsers. Hyperbrowser MCP Server lets Claude, CUA, and other agents scrape, crawl, and extract structured data at scale.
A Mac app that runs local LLMs, Claude, and ChatGPT from one interface. Private by design, works offline, no subscription.
Connect Claude and other AI assistants to a Morphik knowledge base via the official MCP server. Supports multimodal search, file ingestion, and document retrieval.
Open-source toolkit to audit AI agents and MCP servers against NIST AI RMF, OWASP, and GDPR — with a dashboard and 8+ attack testing strategies.
AgentArmor is an open-source Python framework providing 8-layer defense-in-depth security for AI agents, from input ingestion to identity management.
Set up Lantern, the open-source PostgreSQL extension for vector storage and HNSW search. 888 GitHub stars, YC W24, AGPL-3.0 licensed.
Laminar is an open-source observability platform for AI agents. It traces LLM calls, tracks tool use, runs evaluations, and detects hallucinations.
140K+ developers use Metaschool's free AI templates to build apps like flashcards, email generators, and chess games with GPT-4 and Gemini.
Pluely is a privacy-first, open-source AI desktop assistant that stays hidden during meetings and video calls. Built with Tauri and Rust, only 10MB.
DeepReel's AI Video Genie transforms blog posts, PDFs, and plain text into polished explainer videos with AI presenters, voiceovers, captions, and B-roll — in minutes.
Sequel acts as a secure MCP bridge between your databases, SaaS tools, and any AI agent. Connect once, query in plain English across PostgreSQL, BigQuery, Stripe, and 100+ sources.
Try OpenAI's Realtime API voice mode in a browser playground built on LiveKit Agents. No installs, no server setup — just open and talk.
Sipp is an open-source AI runtime that runs GGUF models directly in-browser via WebGPU, with a symmetric API for local execution and cloud gateway routing. Built with Rust and C++.
Reyn is a privacy-first desktop AI for Mac that watches your screen, builds a private work journal, and answers questions — all locally.
Rigour is an open-source governance layer for AI coding agents that scans for secrets, enforces code quality, and controls agent memory in real time.
Open-source web scraping API that converts any URL into clean Markdown or structured data for AI agents. Supports search, scrape, crawl, and map endpoints with LLM-ready output.
Instantly check whether your PC can run open-source LLMs like Llama, Mistral, and Gemma locally. Get personalized GPU/VRAM requirements and quantization recommendations.
Open-source visual programming for backend AI workflows. Build, debug, and iterate agentic AI flows directly in VS Code with full TypeScript integration.
One API to route across OpenAI, Anthropic, Google, and more. Sudo handles context management and monetization so developers can focus on building AI products.
Build AI video editing agents on a node-based canvas with Mosaic — describe the edit in natural language and let multimodal AI handle footage analysis, auto-clipping, AI avatars, and voiceovers.
Mastra is an open-source TypeScript framework for building AI agents and AI-powered applications. From the team behind Gatsby.js, backed by Y Combinator W25.
Build and run AI agentic workflows across channels with YAML definitions, MCP integration, memory support, and multi-channel continuity in one Node.js runtime.
Memobase is a persistent memory server for AI agents that eliminates context silos across Claude, Cursor, and ChatGPT via MCP and OpenAI Actions. Self-hosted option included.
Soma is a Rust-based AI agent and workflow runtime with a TypeScript SDK that brings resumability, MCP integration, and A2A protocol support to your existing AI stack.
Cua provides sandboxes, SDKs, and benchmarks for building AI agents that control full desktops on macOS, Linux, and Windows. MIT-licensed, 18K stars.
CyberCage adds enterprise-grade security to AI coding assistants and MCP servers — discovery, approval workflows, PII detection, and audit logs.
Hopper is an agentic terminal that lets AI agents navigate ISPF, write JCL, debug mainframe jobs, and query VSAM inside IBM z/OS.
Point WebMCP Core at any URL, get AI agent tool definitions in seconds. Auto-crawls forms, APIs, and interactive elements; outputs TypeScript, React hooks, or YAML for Chrome 146+.
Browser Harness connects an LLM directly to your Chrome browser via CDP, letting the agent write its own helper code when tasks require custom logic.
Create full-stack AI apps locally with Dyad, an open-source builder that runs entirely on your machine — no cloud lock-in, bring your own keys, Apache 2.0 licensed.
Open-source document processing tool that uses AI to extract structured JSON data from PDFs, DOCX, and images. Supports custom schemas, multiple file formats, and local LLM integration.
Kinic creates a portable, tamper-proof vector database on the Internet Computer blockchain for AI. Cryptographically secured, user-owned, with zero-knowledge proofs.
RCLI runs a full STT + LLM + TTS pipeline on Apple Silicon with MetalRT acceleration. No cloud, no API keys — just 40 macOS voice actions and sub-200ms latency.
Reverse-engineer any platforms internal APIs by recording browser traffic with Integuru, an open-source AI agent.
Open source platform for exploring, explaining, and evaluating neural network features in models like Gemma and Llama — with a hosted tier for teams.
A practical guide to Nanobot, the open-source standalone MCP host that lets you build dedicated AI agents from any MCP server with a YAML config.
Kryfto is an open-source framework that gives AI agents real web browsing capabilities. Ships a 42-tool MCP server for Claude Code/Cursor/Codex, anti-bot stealth, and Docker-based self-hosting.
An AI browser agent that browses the web in plain English. Telegram bot, REST API, free tier. Automates scraping, bookings, forms, and price monitoring without code.
OpenBrowser lets any LLM control Chrome via natural language, using a CodeAgent that writes and executes Python code against live CDP sessions. 2-6x fewer tokens than Playwright MCP.
Describe what you want and get a production-ready app with auth, database, and deployment wired in from day one. An opinionated AI-first full-stack framework from YC S25.
Kybernis adds an idempotency and execution ledger layer to AI agents. Prevents duplicate refunds, repeated mutations, and inconsistent state. Framework-neutral.
OpenMolt is a Node.js framework for building AI agents with a self-directed reasoning loop, 30+ built-in integrations, and scope-gated security.
Deploy and manage unlimited OpenClaw AI employee instances from a single dashboard. Per-instance access control, unified billing, and volume discounts from $0/mo.
Ubik Studio brings a Cursor-like AI editing experience to local file research. Model-agnostic, human-in-the-loop verification, and a citation engine built to stop hallucinations.
Talk to Claude to build and manage durable workflows. No node editors, no connectors. Zyk deploys on Hatchet with retries, scheduling, and human-in-the-loop built in.
Inconvo is an open-source platform for building data agents that answer natural-language questions over production databases with safe query execution, permissions, and structured outputs.
Open-source platform for background tasks, AI agents, and durable workflows. Postgres-backed, multi-language SDKs, free cloud tier with 100k runs.
TTSLab runs text-to-speech and speech-to-text models entirely in your browser via WebGPU. No server, no API keys, no data sent anywhere.
SigmaShake adds deterministic sub-2ms guardrails to Claude Code, Cursor, Gemini CLI, and Copilot - blocking destructive tool calls before they execute.
Kikubot turns email inboxes into AI agents. Deploy a multi-agent network where agents poll IMAP, run LLM loops, and reply via SMTP - no UI needed.
Eliezer is a ~6K-line TypeScript AI agent that runs locally, edits its own code, and works with any OpenAI-compatible LLM. MIT-licensed.
Production-grade middleware that intercepts AI agent tool calls, evaluates them against policy, and blocks or logs risky actions before execution.
DataFuel API transforms websites into clean markdown for RAG pipelines and LLM training. One endpoint, no scraping code.
Chamber deploys an AI agent named Chambie that monitors, root-causes, and fixes GPU issues across AWS, GCP, Azure, and on-prem clusters — answering in Slack. YC W26.
Spidra is a no-code AI web scraping platform. Point at any URL, describe what you want in plain text, and get structured data back - no CSS selectors, no proxy management.
Vocode is an open-source Python library for building real-time voice-based LLM applications. Deploy to phone calls, Zoom meetings, and more. MIT licensed, 3.7k stars.
Moru gives AI agents their own isolated Linux VM using Firecracker microVMs. Open-source, self-hostable, and works with Claude Code and Codex.
Deploy AI agents into secure cloud VMs with a full REST runtime API. File ops, terminal, exec, and search — controlled entirely from your agent code.
Share and search AI coding sessions across Claude Code, Codex CLI, Gemini CLI, and opencode with this open-source MCP server.
agentmbox gives AI agents their own email inbox without sign-ups or API keys. Agents pay per request via x402 USDC on Solana or Base — $0.50 to create a mailbox, $0.005 to send.
Run Claude Code, OpenCode, Gemini CLI, or any terminal AI on your iPhone or Android in under 60 seconds. Free, E2E encrypted, zero-knowledge architecture.
An analytics workspace built for AI agents. Connect data sources, build dashboards with Claude/Cursor/ChatGPT, and share durable traceable reports instead of one-off chat threads.
RCLI is an on-device voice AI for macOS. A complete STT + LLM + TTS + VLM pipeline running natively on Apple Silicon with sub-200ms latency. No cloud, no API keys.
A local-first AI orchestration system for macOS. Built on a graph architecture with capability-based security, it runs entirely on-device with no cloud dependency.
Actionbook turns any website into an AI-agent-operable page via MCP connector and Chrome extension. No CLI install, handles logins and paywalls, 10x faster than traditional browser agents.
TesterArmy is a YC P26-backed agentic E2E testing platform. Define tests in plain English, let AI agents run them before deployment and in production. No selectors, no scripts.
DroidClaw uses ADB and a perception-reasoning-action loop to let an LLM control your Android phone by tapping, typing, and swiping. No API integrations needed.
Kumbukum is an AGPL-3.0 open-source memory infrastructure that gives AI tools persistent, shared context across sessions, projects, and clients.
Plandex v2 is an open-source AI coding agent designed for long-running, multi-file tasks. It沙盒diff预览、2M上下文、本地Docker部署，适合大型代码库。HN 257pts热榜。
Open source CLI that gives AI coding agents like Cursor and Claude Code persistent memory of your real codebases and documentation.
Sentrial is a YC W26 observability platform that detects silent regressions across AI agent sessions, tool calls, and LLM interactions before end users notice.
A privacy-first, local-first AI meeting notetaker. Transcription runs on your machine, notes save as markdown, and you bring your own LLM. MIT licensed.
Command Center runs multiple AI coding agents in parallel and uses a refactoring agent to turn AI-generated slop into production-ready code. Here's how it works.
Open-source local AI agent with Basna ensemble routing, Agent Forge archetypes, 47 built-in tools, and a web-based Flight Deck dashboard. Supports OpenAI, Claude, Gemini, Ollama, and OpenRouter.
Browser Use is an open-source AI agent framework with 99K GitHub stars that controls real browsers via Python. This guide covers setup with Cursor and Claude Code, CLI usage, and when to pick
Kvlar is an open-source Rust-based policy engine that intercepts AI agent tool calls via MCP, evaluating every action against YAML policies before execution. Fail-closed security with audit logging.
Zoonk is an open-source AI learning platform that turns any topic into structured interactive lessons. Self-host with Next.js, pnpm, and PostgreSQL.
Open-source Agent Alignment Protocol (AAP) makes AI agent behavior observable with Ed25519-signed decision traces, EU AI Act compliance, and cross-provider verification.
Measure how well your hardware runs local AI models. Open-source benchmark tests prompt speed, generation speed, and time-to-first-token for LLMs.
Reality Defender API detects AI-generated deepfakes in images, audio, and video. Enterprise-grade verification for platforms and institutions.
Captain (YC W26) automates RAG pipeline setup for S3, GCS, and Google Drive. Index unstructured data in minutes with multimodal search across docs, images, and sheets.
YC W26 inference API with custom IonAttention engine. Drop-in OpenAI replacement, built for NVIDIA Grace Hopper. Per-second billing, no cold starts.
Dynbox is an intelligent file organizer. Easily sort and find all your files, whether on your cloud storage or local disk.
Plan the prompts, retrieval, model choices, and infrastructure behind an AI workflow. See cost, latency, and dependencies before the architecture turns into code.
One-binary Rust server for self-hosted AI agents with memory, voice, MCP tools, and integrations for Telegram, WhatsApp, Discord, Slack, and more.
Storm MCP connects AI agents to 1,100+ verified MCP servers via one-click deployment. Enterprise security, observability, and zero config required.
Port42 is a native Mac app where AI companions talk, build, and think together—generating tools dynamically from your actual work patterns.
BookmarkOS adds an MCP server to your bookmarks, letting AI tools search and retrieve your X bookmarks by folder, tag, or keyword in real time.
Rowboat is an open-source AI coworker that remembers your projects, codebase, and preferences — like Claude Code but runs on your own infrastructure.
Flowtest.ai is an AI agent that browses your website like a real user, detects issues, and self-heals when layouts change.
A curated directory of 316 tools for building AI agents, LLMs, and RAG pipelines. Browse by category, use case, and stack.
Set up AI agents to build and manage superglue tools via the CLI and superglue skill. Perfect for developers building custom AI integrations.
AgentLair gives AI agents a persistent address, credentials, audit trail, and namespace that survive session restarts. Single-operator indie project out of Norway.
Open-source benchmark that tracks which tools AI models recommend for vibe-coding prompts across beginner to expert scenarios.
A secure persistent AI agent server in Rust. One binary, sandboxed execution, multi-provider LLM support, voice, memory, and integrations for Telegram, WhatsApp, Discord, and more.
Modulus connects code repositories into a unified knowledge graph, helping AI coding agents understand and navigate cross-file context across your entire codebase.
Monitor AI agent costs, logs, and kill runaway bots with this open-source dashboard. Free, open source, self-hosted, takes 2 minutes to set up with Node.js.
CosmicJS just shipped Team Agents – AI teammates that manage your CMS from Slack, WhatsApp, and Telegram. Here's how the integration works and what it means for content teams.
Relvy is a YC F24-backed on-call automation platform that executes runbooks during incidents — reducing MTTR by turning static SOPs into AI-driven response workflows. Free tier available.
ClawsMarket is a marketplace where AI agents discover and run curated software workflows — booking meetings, trading markets, sourcing deals.
Trickle is an AI agentic canvas for vibe coding that lets you build production-ready web apps and sites using natural language, with built-in image and video generation.
An open directory of AI coding tools for Claude Code, Cursor, Replit, and MCP. Discover 70+ curated agents, plugins, rules, and skills in one place.
Twill keeps a persistent dev environment for each repo, forks it for every task, and lets coding agents write back memories, setup fixes, and skills while verifying PRs with tests and screenshots.
AppLaunchFlow uses AI to generate app store screenshots, ASO copy, promo videos, icons, and social graphics from a single workspace — streamlining the entire mobile go-to-market workflow.
A purpose-built LLM for OCR, web scraping, and classification tasks that need consistent, deterministic results — no hallucination, no margin for error.
Caddie is a Slack-based AI agent that connects to your tools, learns your workflow, and executes tasks on demand. Pre-built skills available.
Hexabot is an open-source AI chatbot platform with visual workflows, web and messaging channels, MCP integration, and self-hosted deployment for teams that need privacy and control.
InsForge is a Postgres BaaS that turns natural language prompts into data models, APIs, and backend logic—no SQL or schema migration required.
Rivet is an open-source AI agent development environment built by Ironclad for building real-world AI workflows with visual editing, integrated debugging, and LLM provider flexibility.
TokenSpeed targets agentic LLM inference with static-compiler parallelism, KV-safe scheduling, and aggressive Blackwell-era performance goals in its preview release.
OpenSquilla combines routing, planning, memory, and optional MCP support into a token-efficient local-first agent workflow with practical CLI setup and tradeoffs.
Find matching jobs, create job-specific resumes and cover letters, and apply across 12 ATS platforms with built-in tracking. Free to try.
Engram processes agent experiences into structured long-term memory while Fold provides semantic search across project knowledge. A practical look at building persistent memory for AI agents.
An MCP server that adds a cost signal to AI agents - they plan upfront, stay in budget, and stop when done. Works with any MCP client.
Budibase Agents is an open source internal tools platform with AI agents that automate workflows without code. Self-hosted, visual builder, connects to your existing data sources.
persist.chat is an AI sales agent that finds prospects across 200M+ LinkedIn profiles, X, and email — then writes and follows up across every channel until they reply.
Sentrial monitors your AI agents in production and catches failures, errors, and unexpected behaviors before they reach your users. Built for LLM-based systems.
Superhighway serves live web search to AI agents through the x402 payment protocol, charging per-call in USDC with no API key required.
A curated directory of AI tools for creators, indie developers, and founders. Find tools for video, audio, social media, and more.
Drop Karpathy's viral CLAUDE.md skill into Claude Code or Cursor — 4 behavioral principles that kill over-engineering, drive-by refactors, and assumption bugs.
Capframe maps every tool your AI agents reach, mints scoped capability tokens in Rust, and enforces runtime policy decisions in microseconds—no LLM in the path.
Familiar is an open-source local AI agent that runs entirely on your Mac or iOS device — no cloud, no data leaving your hardware. Build and orchestrate agent workflows privately.
Policy-as-code authorization built for AI agents, with OPA-native design, AuthZEN alignment, and agent-operable access workflows through MCP integration.
Nanobot bridges MCP servers and AI agents, letting you turn any MCP-compatible tool into an autonomous agent with memory, planning, and tool use.
AppLaunchFlow is an AI-powered app store screenshot generator for iOS and Android. Create App Store and Play Store screenshots, ASO copy, and translations in minutes.
ToolJet AI is a collaborative low-code platform for building full-stack enterprise internal tools with AI agents. Sign up for early access.
A cloud-hosted AI agent that builds, deploys, and manages SaaS tools with minimal manual ops. Useful for indie hackers who want an AI teammate across build and launch workflows.
The most comprehensive map of the AI agent ecosystem — 510+ tools across 31 categories including MCP servers, coding agents, and frameworks. Updated daily.
Turn blog posts and docs into professional AI videos with avatars, voiceovers, visuals, and publishing-ready output for content teams that need faster repurposing.
OpenLegion is an AI agent platform for production. Deploy autonomous agents with Docker isolation, vault-proxied credentials, per-agent budgets, and fleet coordination.
Port42 is a native Mac app where multiple AI companions collaborate in shared channels, pulling live data to build interactive visualizations on the fly—no code required.
Port42 watches your actual work patterns and automatically generates tools agents can use. No manual tool registration, no API wrappers—just let it learn how your team operates.
Zatanna reverse-engineers any web portal's network traffic and generates headless APIs that AI agents can execute in seconds. No UI required.
agent-data is a CLI that gives AI agents real-time structured data via a local API - cutting token costs and latency vs browser automation.
Obot is an open-source MCP gateway that manages MCP servers, Skills, access policies, audit logs, and integrations for AI agents. Deploy via Docker, connect to any MCP client.
Visual workflow builder in Rust that treats humans and AI as interchangeable nodes. Built by ex-OpenAI evaluator. 20x faster than Claude Code for complex AI systems.
Langtail is a collaborative prompt management platform that lets teams build, test, and deploy AI prompts with a spreadsheet-like interface, evaluation tools, and an AI firewall.
ZeroEntropy (YC W25) builds embedding models, rerankers, and managed search APIs for RAG pipelines. Uses novel Elo-score training to outperform OpenAI and Cohere on retrieval benchmarks.
Run AI agents in hardware-encrypted TEEs backed by AMD SEV-SNP. Private inference, vTPM-bound keys, and TEE-attested execution for OpenClaw workloads.
Crawl, chunk, and vectorize any website into a queryable knowledge base. REST API with collections, scheduled updates, and built-in RAG for LLM applications.
Augento fine-tunes open-source LLMs for AI agents using reinforcement learning. Drop in a reward function, no dataset required. YC W25, 101pts on Launch HN.
Archil (YC F24) turns any S3 bucket into a POSIX cloud filesystem with sub-millisecond cached reads, mount tokens, and a serverless exec runtime for AI training and agents.
OpenNutrition is a free, ODbL-licensed nutrition database of 300K+ foods, built with a multi-model AI pipeline that ingests USDA, AUSNUT, and CNF and fills gaps with reasoning models.
Construct Computer gives autonomous AI agents a persistent cloud desktop, an email address, and live browser access so they can run real work schedules, not just one-off API calls.
Cq is an open standard for shared agent learning from Mozilla.ai. Local-first knowledge units, FastMCP integration, and team commons for AI coding agents.
Oblivus Cloud offers 18,000+ NVIDIA GPUs from $0.29/hr with transparent per-minute pricing, no hidden fees, and 30-second VM provisioning for ML workloads.
Extract is a YC P25 hosted OCR and document parsing API that turns PDFs, PPTX, DOCX, and scans into text chunks with bounding boxes. 81.9% accuracy, 2x faster than AWS Textract, $3 per 1K pages.
LogClaw is an Apache 2.0 log intelligence platform that ingests logs via OpenTelemetry, runs signal-based anomaly detection, and auto-files Jira tickets through LLMs you control.
anarlog is a free, open-source alternative to Granola. Record, transcribe & summarize meetings locally with your own LLM. MIT-licensed, privacy-first, built with Rust + Tauri.
One (formerly Pica) is a Rust-based open-source agent infrastructure layer that gives any AI agent managed auth, audit logs, MCP access, and 70,000+ tools across 400+ apps via one CLI.
Port42 is a native Mac app where AI companions build interactive UIs and act on your machine through a single API. They ship 40+ releases in 9 days alongside you.
Vela is a YC W26 AI agent that reads your emails, calendars, and communication channels to automatically book meetings across candidates, clients, and time.
TeamOut is an AI agent that plans company retreats entirely through conversation — handling venue sourcing, vendor coordination, flight cost estimation, and.
Omnara is a web and mobile agentic IDE that lets you run Claude Code and Codex sessions from anywhere. Keep your coding agents running even when you're away.
Chroma Cloud brings fully-managed, serverless vector search to production AI apps. Built on Apache 2.0 Chroma Distributed in Rust, it powers agents at Apple.
Vela is a YC W26 AI agent that handles multi-party, multi-channel scheduling across email, SMS, WhatsApp, Slack and phone — booking interviews, following up on.
Tinfoil runs open-source LLMs on cloud GPUs with hardware-backed privacy guarantees. Your data never leaves the secure enclave—not even Tinfoil can access it.
Livedocs brings AI-powered data analysis into documents without code. Connect live data sources, ask questions in plain English, and get instant tables and.
Sentrial monitors AI agents in production, detecting loops, hallucinations, and tool misuse the moment they happen. Here's how it works and how to set it up.
RunRL lets developers improve AI models and agents using reinforcement learning by simply defining a reward function — no GPU cluster management required.
CloudRouter is a skill and CLI that lets AI coding agents like Claude Code spin up cloud VMs with GPUs on demand, run dev servers and browser automation, then.
Plexe is a YC X25-backed platform that turns natural language ML problem descriptions into deployed production models — automating feature engineering.
Dedalus Labs spins up persistent full Linux VMs for AI agents in under 250ms. No Dockerfiles, no YAML configs — just SSH, GPU, and a Python SDK.
Nia indexes your codebases, docs, and wikis into a retrievable context layer that stops AI coding agents from hallucinating outdated API references and wrong.
Dedalus Labs connects any LLM to any MCP tool via a single OpenAI-compatible API endpoint. Upload MCP servers to the cloud, skip the Dockerfiles and YAML.
Captain automates building and maintaining RAG pipelines for unstructured data. Index S3, GCS, and Google Drive in minutes with one API call.
Chamber is a YC W26 AI teammate that autonomously monitors, root-causes, and remediates GPU infrastructure issues across multi-cloud.
Extend is a YC W23 AI toolkit that parses messy documents—PDFs, images, Excel files—and extracts reliable structured data for AI pipelines.
Spine Swarm runs parallel AI agents across 300+ models simultaneously, enabling research and client-ready deliverable creation through a visual canvas.
Cloud browsers built for AI agents that need to navigate, scrape, and interact with the web at scale. Handles proxies, stealth detection, and concurrent.
Prism is a YC X25 AI video platform combining generation, editing, and an API for workflow automation. Generate assets, edit on a timeline, and integrate via.
IonRouter is a YC W26 inference API routing open-source and fine-tuned models via an OpenAI-compatible endpoint, built on a C++ runtime optimized for GH200.
YC W26-backed AI agent observability platform. Trace sessions, detect silent regressions, and A/B test prompts in production before failures reach users.
Recall gives Claude Code a permanent memory store that survives session restarts and context compaction. Four hooks capture and restore context automatically — with cloud SaaS or self-hosted options.