Welcome to the LangGraph package reference documentation!
LangGraph is a framework for building resilient language agents as graphs. The langgraph package is the main entrypoint, providing everything you need to define stateful, multi-step agent workflows. The packages below form the LangGraph ecosystem:
langgraphThe core framework for building stateful, multi-actor applications with LLMs as graphs.
langgraph.prebuiltPre-built agents and tool nodes for common patterns like ReAct agents.
langgraph.checkpointBase interfaces and serialization for graph state checkpointing.
langgraph.checkpoint.postgresPostgreSQL-backed checkpoint storage for persistent graph state.
langgraph.checkpoint.sqliteSQLite-backed checkpoint storage for lightweight local persistence.
langgraph_cliCommand-line interface for managing and deploying LangGraph applications.
langgraph_sdkPython SDK client for interacting with deployed LangGraph APIs.
langgraph_supervisorBuild multi-agent systems with a supervisor agent that delegates to sub-agents.
langgraph_swarmBuild multi-agent systems with swarm-style agent handoff patterns.