Skip to main content
AgentsUse

Python agent framework

LangGraph

A low level orchestration framework for building, managing and deploying long running stateful agents.

Official siteDocumentationRepository
PythonMITChecked 2026-10-07Source verified

Quick decision

Ecosystem
Python, TypeScript
License
Open source, MIT
Package
PyPI: langgraph, version 1.2.14
Review state
Source verified. Facts checked 2026-10-07. Not locally tested by AgentsUse.

What it is

LangGraph is described on PyPI and GitHub as a low level orchestration framework for stateful agents. It provides infrastructure for durable execution, streaming, human in the loop, persistence and memory, and is recommended when workflows need deterministic and agentic steps combined with careful control.

The repository highlights durable execution that resumes after failure, human oversight through inspecting and modifying state, comprehensive short term and long term memory, debugging with LangSmith, and production ready deployment for stateful workflows.

LangGraph can be used standalone or with the LangChain ecosystem. The repository notes it is built by LangChain Inc, pairs with LangChain integrations and LangSmith, and has an equivalent JavaScript library, while LangChain agents themselves are built on top of LangGraph.

What it does

  • Provides low level graph based orchestration for long running stateful agents and workflows.
  • Supports durable execution with checkpointing and resumption after failures.
  • Supports human in the loop inspection and modification of agent state.
  • Provides short term working memory and long term persistent memory primitives.
  • Integrates with LangSmith for tracing execution paths, state transitions and runtime metrics.

What it does not do

  • It is not a high level prebuilt agent product by itself; the repository points quick agent builders to higher level packages and LangChain.
  • It does not supply search, browser or scraping tools directly; those are supplied through LangChain integrations or MCP servers.
  • It does not replace a deployment platform or observability setup for production stateful workflows.

Setup

Install LangGraph

The repository quickstart installs the Python package with pip, and the docs link to a LangGraph quickstart for building the first graph.

Install LangGraph
pip install -U langgraph

Environment variables (names only)

  • OPENAI_API_KEY
  • LANGSMITH_API_KEY

Source: https://github.com/langchain-ai/langgraph. Examples use placeholders only.

MCP support: LangGraph agents can consume MCP tools through langchain-mcp-adapters, and LangChain documentation shows MCP tools loaded into agents that can run on LangGraph.

Health and maintenance

GitHub stars

42,851

Checked 2026-10-07

GitHub forks

7,284

Checked 2026-10-07

Maintainer
LangChain Inc

Stars and forks read from the GitHub repository page in this run. PyPI page shows release files for langgraph 1.2.14 dated Oct 6, 2026.

Verified compatible tools

LangSmithBuilt in by the framework vendor

Via LangSmith debugging and observability for LangGraph. The LangGraph repository README lists Debugging with LangSmith and pairs LangGraph with LangSmith for agent evals and observability. Evidence. Setup detail

ExaOfficial adapter package

Via langchain-exa ExaSearchResults and ExaFindSimilarResults used with a LangGraph agent. The LangChain Exa search page states the Exa tools can be used with a LangGraph agent to dynamically search and find similar content. Evidence. Setup detail

TavilyOfficial adapter package

Via langchain-tavily TavilySearch used as a LangChain tool in LangGraph agents. The Tavily provider page states langchain-tavily exposes Tavily endpoints as LangChain tools; LangGraph agents consume LangChain tools, and the LangGraph repository states LangChain provides integrations and composable components. Evidence. Setup detail

PlaywrightConnected through MCP

Via langchain-mcp-adapters (MultiServerMCPClient, load_mcp_tools). The fetched LangChain docs mirror shows pip install langchain-mcp-adapters and load_mcp_tools used to load remote MCP tools into an agent, which is the documented MCP path for LangChain and LangGraph agents. Evidence. Setup detail

Grouped view: tools for LangGraph.

Stacks featuring this framework

Research Briefing Stack

Tavily finds and extracts sources, a LangGraph loop plans and drafts, and Langfuse keeps every run inspectable.

Sources and freshness

Last checked 2026-10-07. Verification label: source verified. It does not mean AgentsUse ran the framework. Spotted an error? Send a correction. Back to the framework index.