Catalog

Lightweight framework for building multi-agent AI workflows. Provider-agnostic, supports OpenAI Responses API, Chat Completions API, and 100+ LLMs. Features sandbox execution, human-in-the-loop, tool guardrails, and built-in tracing.

Open-source Python framework for building LLM applications and agents. A uv monorepo of independently-versioned packages: core primitives, the v1 agent framework, legacy classic package, text splitters, standard tests, model profiles, and 15+ partner integrations.

Self-hosted, multi-channel gateway for AI agents. Run one Gateway on your own hardware and message an always-available AI assistant from Discord, Slack, Telegram, WhatsApp, Signal, iMessage, Matrix, and more. Agent-native (tool use, sessions, memory, multi-agent routing), plugin-heavy, terminal-first, MIT-licensed. Written in TypeScript for hackability.

The AI browser agent. Open-source async-Python (>=3.11) library that drives a real Chromium browser over the Chrome DevTools Protocol (CDP), letting LLM agents navigate pages, read the DOM, and act autonomously. PyPI: browser-use. Docs: docs.browser-use.com.

Open-source AI Gateway and Python SDK giving a single, unified OpenAI-format interface to 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and more). Use it as a Python SDK for direct integration, or deploy the AI Gateway (Proxy Server) as a centralized service with virtual keys, spend tracking, budgets, guardrails, load balancing, and an admin dashboard. MIT-licensed, by BerriAI (YC W23). PyPI: litellm; docs: docs.litellm.ai.

Lean, fast Python framework for building autonomous multi-agent AI systems, independent of LangChain. Offers two paradigms: Crews (autonomous agent teams) and Flows (event-driven orchestration).

Comprehensive skills library for Claude AI and Claude Code -- 388 production-ready skill packages bundling domain expertise, Python tools (727), reference guides (842), and strategic frameworks across 20 domains. Distributed as 99 marketplace plugins, with 118 agents and 150 slash commands. (v2.12.0 consolidated release + post-tag unreleased work: agent-launcher CMA domain; engineering plugins agent-memory, deep-learning-book, human-gate, skill-doctor, spinning-up-deep-rl; marketing/linkedin; productivity/swedish-mentor; research/deepread.)