Agent framework shortlist
Updated 2026-08-24

Best AI agent frameworks on GitHub

Libraries you import to build agents. Not chat UIs, not MCP servers, not SKILL.md packs. One strong option per style: chains, graphs, multi-agent, typed Python, first-party SDKs.

How we pick

  • Must be a library/runtime for building agents, not a hosted chatbot product.
  • Public GitHub repo with English docs.
  • One representative per style: chains, graphs, multi-agent, typed Python, first-party SDKs.
  • Skip prompt packs and SKILL.md collections (those live on the skills page).
  • Skip self-hosted apps like Open WebUI (those live on the self-hosted page).

Who this is for

English-speaking builders choosing how to orchestrate models and tools. If you want something you can docker-compose, use the self-hosted list instead.

The shortlist

Icon, name, one sentence. Open a card for stars and the repo detail page.

#1

langchain

Orchestration

The default Python/JS toolkit for chaining models, tools, and retrieval.

Builders 149,870 stars

Still the comparison point for “how do I wire an LLM to tools” in English search.

#2

langgraph

Graph agents

Stateful, graph-based agent workflows on top of the LangChain ecosystem.

Builders 45,123 stars

The production-shaped successor when a linear chain is not enough.

#3

crewAI

Multi-agent

Role-based multi-agent crews for research, writing, and ops-style jobs.

Team 61,083 stars

The most-searched multi-agent productized framework after LangChain.

#4

autogen

Multi-agent

Microsoft AutoGen for multi-agent conversations and tool-using workflows.

Builders 61,929 stars

Enterprise-familiar name; useful as the Microsoft-stack alternative on the list.

#5

pydantic-ai

Typed agents

Pydantic-first agents: structured outputs, tools, and type-safe Python.

Builders 21,278 stars

Fills the “I want agents that look like normal Python” slot.

#6

openai-agents-python

OpenAI agents

OpenAI’s official Python agents SDK for tool calling and handoffs.

Builders 30,753 stars

First-party SDK for teams already standardized on OpenAI APIs.

#7

smolagents

Code agents

Small Hugging Face agents that write and run code to finish a task.

Builders 30,433 stars

Lightweight alternative to heavy orchestration graphs.

#8

adk-python

Google ADK

Google Agent Development Kit for building agents on Gemini and Vertex.

Team 22,263 stars

The Google-cloud counterpart so the list is not only OpenAI/LangChain.

#9

agno

Agent OS

Agno (formerly Phidata) for fast agents with tools, memory, and teams.

Builders 43,234 stars

High-velocity Python agent runtime that shows up in “faster than LangChain” searches.

#10

llama_index

RAG agents

LlamaIndex for data-connected agents and retrieval over private documents.

Team 52,946 stars

The RAG-native framework when the job is “talk to my files,” not generic chat.

FAQ

How is this different from self-hosted AI tools?

Self-hosted AI is products you run (Open WebUI, Dify, Ollama). This page is frameworks you import to build your own agent. Different search intent, different shortlist.

Do I need LangChain?

Not automatically. Use LangGraph or LlamaIndex if you already know the job is stateful workflows or RAG. Use PydanticAI or the OpenAI Agents SDK if you want less framework and more typed Python.

Where are MCP and skills?

MCP servers give agents tools. Skills give agents procedures. Frameworks glue models, tools, and state. They are three separate pages on purpose.

How often is this updated?

Weekly. Renames (for example Phidata → Agno) are updated on the next pass.