Often times that Claude code just took an approach where it’s not recommended by LangGraph even with the LangGraph MCP connected. What I do now is I cross-check with the “Chat LangChain" in the docs to verify if Claude is doing something right or not.
I wonder if there is a way for Claude code to just talk to this chat agent?
Hi @wanchaiS, welcome to the LangChain community!
What I’ve done in Cursor is create a set of rules that allow it to act as a “LangChain expert,” highlighting best practices. In those rules, I’ve specified that it should always consult the LangChain documentation (already connected Langchain Docs MCP in cursor) before taking any steps.
Unfortunately, I can’t directly share those rules right now, as they’re part of a private repository for the company I’m working with. However, you could create a specialized subagent in Claude Code replicating a similar approach as my rules.
As for your question, to the best of my knowledge, there’s currently no way to connect Claude Code with LangChain’s AI Docs Agent.
1 Like
Since every LangSmith deployment exposes an MCP endpoint, you can add it:
claude mcp add chat-langchain --transport http https://chat-langchain-993a2fee078256ab879993a971197820.us.langgraph.app/mcp
1 Like