LangChain vs LangGraph
In 1,621 Python projects set up for AI coding agents, LangChain is a dependency in 126 (7.8%) and LangGraph in 77 (4.8%): LangChain is 1.6× as common. Among repos created in the last 90 days that use a agent, mcp and tool libraries tool, LangChain is in 4.9% and LangGraph in 2.4%, against 17.7% and 10.8% of older ones. 67 projects list both, often a sign of a migration or of a monorepo with several apps.
Agent, MCP and tool libraries PyPI Updated Nov 17, 2026
Side by side
| LangChain | LangGraph | |
|---|---|---|
| Agent projects that use it | 126 | 77 |
| Rank in agent, mcp and tool libraries | #4 | #5 |
| In new repos with a agent, mcp and tool libraries tool (last 90 days) | 4.9% | 2.4% |
| In older repos with one | 17.7% | 10.8% |
| Downloads, last 7 days | 36.5M | 14.6M |
| GitHub stars | 150K | 45.1K |
| Last push to its repo | Nov 17, 2026 | Nov 17, 2026 |
| License | MIT | MIT |
| Funding | — | — |
| Projects that list both | 67 | |
Downloads include CI runs and installs as a dependency of other packages. Stars and pushes are from GitHub's API; downloads from PyPI.
By coding agent
| Projects set up for | Python repos | LangChain | LangGraph |
|---|---|---|---|
| Claude Code | 782 | 5.4% | 3.8% |
| GitHub Copilot | 805 | 9.2% | 5.7% |
| Cursor | 548 | 9.5% | 6.0% |
| Gemini CLI | 83 | 13.3% | 0% |
| OpenAI Codex | 78 | 7.7% | 0% |
| OpenCode | 53 | 9.4% | 0% |
Share of the Python repos configured for each agent (its config file in the repo) that list the tool. A repo set up for two agents counts for both.
LangChain
Framework for composing applications with LLMs through chains and agents.
Most often used with it:
| python-dotenv | 71.4% of its projects |
|---|---|
| Pydantic | 57.1% of its projects |
| FastAPI | 53.2% of its projects |
| LangGraph | 53.2% of its projects |
| Uvicorn | 52.4% of its projects |
pip install langchain
LangGraph
Framework for building stateful, multi-actor applications with LLMs.
Most often used with it:
| LangChain | 87.0% of its projects |
|---|---|
| Pydantic | 72.7% of its projects |
| python-dotenv | 71.4% of its projects |
| FastAPI | 64.9% of its projects |
| Uvicorn | 63.6% of its projects |
pip install langgraph
How we count
Projects are public GitHub repos our weekly scan finds through coding-agent config files (CLAUDE.md, AGENTS.md, .cursor/mcp.json and others); a project uses a tool when its package.json, requirements.txt or pyproject.toml lists it. "Same job" was checked for every pair, so add-ons and companions are never paired. Shares by repo age compare repos created in the last 90 days with older ones in the same scan. More on the method.