Awesome Open-Source AI

A curated list of open-source AI projects worth reading, running, and contributing to, with an emphasis on agentic AI, LLM tooling, and the Model Context Protocol (MCP).

Maintained by @opensource-joe as a learning-in-public map of the open-source AI ecosystem. Each entry is a project I am reading through, running, or contributing to. Suggestions and corrections are welcome.

Contents

• Agentic frameworks
• Agentic coding tools
• Model Context Protocol (MCP)
• LLM tooling and CLIs
• Inference and serving
• Learning resources

Agentic frameworks

Libraries for building multi-step, tool-using AI agents. Good for learning agent design patterns.

• LangChain - The most widely used framework for chaining LLM calls, tools, and memory (Python).
• LlamaIndex - Data framework for connecting LLMs to your own data, RAG (Python).
• AutoGen - Microsoft's framework for multi-agent conversations and workflows (Python).
• CrewAI - Role-based orchestration of collaborating agents (Python).
• DSPy - Program, rather than prompt, language models; compiles prompts and weights (Python).
• Letta - Stateful agents with long-term memory, formerly MemGPT (Python).

Agentic coding tools

AI agents that read, write, and run code. The closest analog to tools like Claude Code.

• OpenHands - Open platform for AI software-development agents; active, beginner-friendly issue tracker (TypeScript/Python).
• Aider - AI pair programming in your terminal, tightly integrated with Git (Python).
• SWE-agent - Agent that resolves real GitHub issues, from the SWE-bench team (Python).
• Goose - Extensible local AI agent for engineering tasks (Rust).

Model Context Protocol (MCP)

The open standard for connecting AI models to tools and data. A high-leverage, fast-moving area to learn and contribute.

• MCP Servers - Reference implementations of MCP servers (TypeScript).
• Python SDK - Build MCP clients and servers in Python (Python).
• TypeScript SDK - Build MCP clients and servers in TypeScript (TypeScript).

LLM tooling and CLIs

Practical tools for working with LLMs day to day. A friendly on-ramp for first contributions.

• llm - CLI and Python library for running prompts across models, with a rich plugin ecosystem (Python).
• Datasette - Explore and publish data; growing set of LLM/AI plugins (Python).

Inference and serving

Run and serve open models locally or at scale.

• Ollama - Run open LLMs locally with a simple CLI and API (Go).
• vLLM - High-throughput, memory-efficient inference and serving engine (Python).
• llama.cpp - LLM inference in plain C/C++, the backbone of much local AI (C++).
• Transformers - Hugging Face's library of model definitions and pipelines (Python).

Learning resources

Docs, notebooks, and courses for learning by doing (and easy first PRs, typos and broken examples included).

• Anthropic Cookbooks - Runnable recipes for building with Claude (Jupyter).
• Anthropic Courses - Structured courses on prompting, tool use, and more (Jupyter).
• Anthropic SDK for Python - Official Python client for the Claude API (Python).

Contributing

Found a great open-source AI project, a better description, or a broken link? Please open a pull request or an issue. See CONTRIBUTING.md for the entry format and guidelines.