LangGraph Overview
Status: ACTIVE (pulled from docs.langchain.com) Source: https://docs.langchain.com/oss/python/langgraph/overview Timestamp: 2026-05-11
What is LangGraph
LangGraph is a low-level orchestration framework and runtime for building, managing, and deploying long-running, stateful agents. It focuses entirely on agent orchestration. Trusted by Klarna, Uber, J.P. Morgan, Replit, Elastic, and more.
Core Benefits
- Durable execution: Agents persist through failures and resume from where they left off.
- Human-in-the-loop: Inspect and modify agent state at any point during execution.
- Comprehensive memory: Short-term working memory and long-term persistent memory across sessions.
- Debugging with LangSmith: Visualize traces, capture state transitions, runtime metrics.
- Production-ready deployment: Scalable infrastructure for stateful, long-running workflows.
Core Concepts
| Concept | What It Is |
|---|---|
| StateGraph | The main graph class. Parameterized by user-defined State object. |
| Nodes | Python functions that encode agent logic. Receive current state, return updates. |
| Edges | Deterministic or conditional routing between nodes. |
| State | Shared data structure. TypedDict, dataclass, or Pydantic model with reducers. |
| Checkpointer | Persists state at each super-step, enabling durability and human-in-the-loop. |
| Command | Combine state updates with routing in a single return value. |
| Send | Dynamically fan out to parallel node executions (map-reduce). |
Two APIs
| API | Best For | Mental Model |
|---|---|---|
| Graph API (StateGraph) | Fine-grained topology control | Build a graph of nodes and edges |
| Functional API (@entrypoint) | Adding persistence to existing code | Write a single function with @task decorators |
How LangChain Products Fit Together
| Layer | Product | What It Provides |
|---|---|---|
| Harness | Deep Agents SDK | Planning, subagents, filesystem, context management |
| Framework | LangChain | Abstractions for models, tools, agent loops |
| Runtime | LangGraph | Durable execution, streaming, HITL, persistence |
| Platform | LangSmith | Tracing, evaluation, deployment, monitoring |
Installation
pip install -U langgraph
Acknowledgements
LangGraph is inspired by Pregel and Apache Beam. The public interface draws inspiration from NetworkX. Built by LangChain Inc but can be used without LangChain.