OSS TanbouOn this page
Overview
LangGraph is a Python framework for building and managing long-running, stateful AI agents and workflows as graphs. It provides low-level facilities for durable execution, streaming, memory, persistence, and human-in-the-loop control. It fits systems that mix deterministic processing with agent decisions and need explicit control over state transitions, pauses, and recovery.
Features and best fit
Based on official documentation; not hands-on tested · Content checked:
Key features
Model state and node transitions explicitly as a graph
Developers define nodes that read or update shared state and connect them with edges, including sequential paths, branches, loops, and subgraphs. Nodes are not limited to model calls; deterministic functions and external tools can participate in the same workflow.
Sources: [1]
Resume long-running work from checkpoints and insert human decisions
Durable execution supports recovery after failures or interruptions, while short-term and long-term memory can be combined with persistence. Interrupts let an operator inspect or modify state before continuing, providing supervision for long-running agents.
Sources: [1]
Best fit
Fits agents that need explicit state, recovery, and approval paths
LangGraph suits workflows that mix deterministic steps with model decisions, resume after failure, wait for approval, or retain conversational and business state across sessions. When the goal is simply to assemble a prebuilt high-level agent quickly, LangChain or Deep Agents may require less application code.
Sources: [1]
Before adoption
Low-level control leaves state, persistence, and side-effect design to the application
The application must define graph structure, state schemas and reducers, checkpoint storage, retry idempotency, and tool permissions. The framework does not guarantee model or tool quality, cost, availability, or rate limits. Python and JavaScript/TypeScript implementations live in separate repositories, and packages across the ecosystem version independently, so lockfiles and regression evaluations remain important. LangGraph uses the MIT License.
Official sources
- [1]LangGraph README at observed commit(2026-10-04)
- [2]LangGraph package README at observed commit(2026-10-04)
- [3]LangGraph MIT license at observed commit(2026-10-04)
Supplemental curator note
Evaluate LangGraph as an execution substrate for explicit state transitions and recovery, not as the quickest set of high-level agent building blocks. Nodes that call external systems should be idempotent because checkpoint recovery can repeat work.
Try it in 3 steps
- 1
Create a virtual environment
Keep the trial dependencies separate from an existing Python environment.
python -m venv .venv . .venv/bin/activate - 2
Install LangGraph
Add the Python package documented in the official README to the virtual environment.
python -m pip install -U langgraph - 3
Run a minimal graph without a model
Confirm that execution moves from START through one node to END and updates state. No API key is required.
python -c 'from typing_extensions import TypedDict; from langgraph.graph import StateGraph, START, END; State=TypedDict("State", {"message": str}); g=StateGraph(State); g.add_node("hello", lambda s: {"message": s["message"]+" LangGraph"}); g.add_edge(START, "hello"); g.add_edge("hello", END); print(g.compile().invoke({"message": "Hello"}))'
Growth
Growth trends · Last 30 days
42,696 Stars
Trend data is still being collected.
Development activity
Last 90 days · weekly
- Commits (last 30 days)
- 56
- Open PRs
- 234
Development activity is still being collected.
Built with
Categories and tags
GitHub data
GitHub dataView detailed GitHub data
GitHub Topics
- agents
- ai
- ai-agents
- chatgpt
- deepagents
- enterprise
- framework
- gemini
- generative-ai
- langchain
- langgraph
- llm
Related information
Write a related articleShare a guide or use case for this OSS in Markdown. Articles are published after administrator approval.
Explore next
- Agno42,541 Stars
3 shared tag(s) · 2 shared category(s) · Same language
treat agent development and the production runtime as one platform stack
Python - smolagents29,667 Stars
3 shared tag(s) · 2 shared category(s) · Same language
Build code-writing AI agents with a compact Python library and pluggable models, tools, and execution backends
Python - SWE-agent20,486 Stars
3 shared tag(s) · 2 shared category(s) · Same language
a research software-engineering agent that works from issues to repository patches
Python - TestZeus Hercules1,174 Stars
3 shared tag(s) · 2 shared category(s) · Same language
turn Gherkin scenarios into LLM-agent browser actions for end-to-end testing
Python - Flowise55,487 Stars
3 shared tag(s) · 2 shared category(s)
treat the visual AI workflow platform as an EOL migration and maintenance target
TypeScript - FastGPT29,780 Stars
3 shared tag(s) · 2 shared category(s)
assemble document-grounded answers and AI workflows visually
TypeScript
Report incorrect information
Tell us if any listing information is incorrect or outdated.