LangGraph

LangGraph is an open-source framework from LangChain for building stateful AI agents as graphs, with memory, streaming and human approval steps.

LangGraph is a low-level orchestration framework and runtime for developers who need agents that handle long, multi-step tasks reliably. Workflows are written as graphs of nodes and edges over shared state, which makes single-agent, multi-agent and hierarchical designs possible in one framework. Built-in persistence saves state as the agent runs, so it can pause for a person to review or approve an action, resume later and keep memory across sessions. Native streaming sends tokens and intermediate steps to the interface as the agent works. It is MIT-licensed, free to use, works with any model provider and is available for Python and JavaScript. LangChain, which maintains it, released version 1.0 in October 2025 and offers LangSmith for tracing, evaluating and deploying the agents built with it.

LangGraph at a glance

Best for: Software engineer
Runs on: self-hosted, open source
Pricing: Free and open source under the MIT licence
Released: January 2024
Category: AI agent tool
Website: langchain.com

Key features

Graph-based workflows with nodes, edges and shared state
Human-in-the-loop pauses for review and approval
Built-in persistence and memory across sessions
Token-by-token streaming of agent steps
Single-agent, multi-agent and hierarchical control flows
Python and JavaScript libraries under the MIT licence

Use it for

Build a support agent that asks a person before issuing refunds
Coordinate several specialist agents on one research task
Run long tasks that resume after a pause or failure

Companions

Tools that articles about LangGraph recommend using with it, each with its job.

LangChain — Built to work with LangChain.
LangSmith — Step-by-step traces and evals that show what an agent actually did
Pinecone — Managed retrieval for the fastest path to production.
Qdrant — Self-hosted vectors when cost or data residency matter.
Ollama — Fully self-hosted inference paired with Qdrant.
Langfuse — Self-hostable alternative to LangSmith.

Alternatives to LangGraph

Langflow — Builds and deploys AI agents visually with Python and major LLMs.
Langroid — Simplifies multi-agent orchestration and app development with intuitive abstractions.
CodeGraph — It eliminates file crawling, reducing AI agent tool calls by 88% and cutting costs.
LangWatch — Runs simulations and continuous evaluations to make AI agents reliable.
AG-UI — Standardizes agent-user interactions and generative UI for applications.
GraphJin — Provides a governed, verified AI agent layer over existing data and systems.
LangChain4j — Integrates LLMs and vector stores directly into Java applications.
Graphlit — Continuous ingestion, entity resolution, and provenance via a single API.
Graphiti — Provides incremental updates and hybrid retrieval for knowledge graphs.
AGiXT — Builds autonomous agents with integrated memory and web capabilities.
TrustGraph — Provides ontology-grounded knowledge for auditable and compliant AI workloads.
Flowise — Modular building blocks and visual tools to build and scale agentic systems fast.

All LangGraph alternatives