Not a simple GitHub list. We've hand-picked the most worthy projects from thousands, categorized by class, maturity, and learning curve. Each project has a core explanation and learning path suggestion.
Filter by category, tech stack, or maturity. Search directly by name or description.
Industry-standard framework: chain composition, tool integration, memory, RAG all in one. Richest ecosystem.
Abstract Agent as directed graph: state checkpoint, human-in-the-loop, loop branches. Modern Agent construction.
Role-based multi-Agent framework. Each Agent has role, goal, tools, backstory.
Microsoft's dialogue-driven multi-Agent: group chat, human-in-loop, termination conditions.
Data ingestion and RAG framework: rich index structures, complex data source integration.
OpenAI's official SDK (upgrade from Swarm): Handoffs, Guardrails, Tracing native.
HF minimal Agent library, perfect for learning Agent fundamentals.
Stanford's revolutionary "prompt compiler". Program LLM pipelines, auto-optimize prompts and models.
RAG-specific eval framework with faithfulness, answer relevance, context precision metrics.
Observability for Agent systems. LangSmith commercial, LangFuse OSS, both provide trace, eval, prompt mgmt.
Microsoft's .NET ecosystem LLM orchestration framework. Native integration with enterprise systems.
LangChain's visual drag-drop interface. Non-developers can also build Agents.
Revolutionary hierarchical memory: helps LLMs break context window limits with true long-term memory. Now renamed to Letta.
Abstract multi-Agent as software company: PM, architect, engineer, QA collaborative flow.
Open-source Devin: Agent-driven software engineer that resolves GitHub issues, writes code, tests, debugs.
Featured 16 projects ยท More curated projects on Resources
Not the newest or the most popular โ the right one for your specific scenario.
| Scenario | First Choice | Alternative | Key Consideration |
|---|---|---|---|
| General Agent apps | LangGraph | OpenAI Agents SDK / Pydantic AI | Stateful, debuggable, production-trusted |
| Multi-Agent role collaboration | CrewAI | AutoGen / MetaGPT | Intuitive API, clear roles |
| Dialogue/multi-Agent group chat | AutoGen | CrewAI | Termination conditions, observable dialogue |
| RAG-dominant apps | LlamaIndex | LangChain + independent Agent layer | Rich indexes, retrieval strategies |
| OpenAI ecosystem integration | OpenAI Agents SDK | LangChain OpenAI integration | Native Tracing, Handoffs |
| Learning / education | Smolagents | BabyAGI | Few lines, easy to read |
| Prompt auto-optimization | DSPy | Manual iteration | Academically rigorous, quantifiable |