Skip to content
WhichAgent

LangGraph

Agent framework by LangChain, Inc.

ActiveOpen source · MITVerified Oct 22, 2026
Website Docs Source

Low-level open-source orchestration framework and runtime for long-running, stateful agents modeled as graphs, with durable execution, streaming, and human-in-the-loop support.

Key facts

Type
Agent framework · Library
Pricing
Free tier, paid from $39/mo
Free tier
The LangGraph library is free and MIT-licensed. Managed hosting is via LangSmith Deployment (formerly LangGraph Platform); the free LangSmith Developer plan does not include deployments, Plus includes one small serverless deployment.
License
MIT
Models
Model-agnostic (any LLM callable from Python/JS), OpenAI, Anthropic, Google, Local models via Ollama, All providers available through LangChain integrations
Languages
Python, TypeScript
Platforms
macOS, Windows, Linux
Bring your own API key
Yes
Local models
Yes
MCP support
Yes
Self-hostable
Yes
GitHub stars
43,714
Last commit
Oct 21, 2026

LangGraph pricing

LangSmith Developer

$0/month

1 seat, 5k base traces/month; no deployments included

LangSmith Plus

$39/month

Per seat per month, billed monthly; 1 free Serverless (Small) deployment; additional Serverless and Dedicated deployments metered (runtime 0.0675 LSU/vCPU-hr, 0.0090 LSU/GiB-hr; 1 LSU = $1); no yearly price listed

LangSmith Enterprise

Custom

Contact sales; custom deployment terms; cloud, hybrid, or self-hosted

Best for

  • Agents that need explicit control flow mixing deterministic steps and LLM calls
  • Long-running workflows that must persist state and resume after failures
  • Human-in-the-loop approval or editing of agent state

Pros

  • Fine-grained control over agent state and control flow
  • Built-in persistence, durable execution, and short- and long-term memory
  • Usable without LangChain, and available for Python and JavaScript/TypeScript
  • Optional managed deployment via LangSmith Deployment

Cons

  • Lower-level than prebuilt agent frameworks, so more code is needed for simple agents
  • Graph/state concepts add a learning curve
  • Managed deployment and Studio tooling are tied to the paid LangSmith platform

How LangGraph compares

CrewAI vs LangGraph

Pick LangGraph when you need explicit, low-level control over every step: shared state, branching and loops, with checkpointing and human interrupts available at any node.

Updated Oct 22, 2026

LangChain vs LangGraph

They aren't rivals: LangChain's agents are built on top of LangGraph.

Updated Oct 22, 2026

Sources (7)