LangGraph

LangChain’s low-level orchestration framework and runtime for long-running, stateful agents, modelled as graphs mixing deterministic steps and LLM decisions.

• Status: LangGraph 1.0 and LangChain 1.0 reached GA together on 2025-10-22, the first stable majors, with a “no breaking changes until 2.0” commitment (LangChain blog). Python and TypeScript.
• LangChain 1.0 adds create_agent (built on the LangGraph runtime), a middleware system and standardized content blocks; legacy features moved to langchain-classic.
• Key concepts: durable execution with checkpoints, persistence, human-in-the-loop state inspection, short- and long-term memory, streaming.
• Ecosystem: LangSmith for tracing, evals and deployment.
• Protocols: AG-UI originated with LangChain and CopilotKit partnership (ag-ui); A2A compatibility reported by Google (a2a-protocol); MCP via adapters (not verified this session, see model-context-protocol).
• Compare crewai, mastra, microsoft-agent-framework.

Sources

• https://www.langchain.com/blog/langchain-langgraph-1dot0 (2026-09-30)
• https://docs.langchain.com/oss/python/langgraph/overview (2026-09-30)