LangChain

LangChain provides the foundational components for connecting LLMs to tools and data sources. LangGraph is a separate orchestration framework built on top of LangChain that enables stateful, multi-step agent workflows using a graph-based architecture with nodes, edges, and persistent memory.

LangSmith operates as an independent observability platform. It can trace and monitor any AI application through its SDK and API, regardless of whether the application uses LangChain or LangGraph frameworks. It supports Python, JavaScript, and REST-based instrumentation.

Deep Agents is a higher-level package built on LangChain designed for autonomous, long-horizon tasks. It includes built-in capabilities for planning, subagent coordination, file system usage, and conversation history summarization, supporting pluggable sandbox environments for code execution.

Yes. LangSmith Deployment supports exposing deployed agents as Model Context Protocol servers. This allows other AI applications and agents to discover and invoke your deployed agent as a tool, enabling inter-agent communication through a standardized protocol.

LangGraph provides built-in checkpointing with PostgreSQL persistence. Agent state is saved at each node execution, enabling conversations to resume after server restarts, human-in-the-loop interrupts, and parallel branch exploration through snapshot forking.