Unexpected langgraph-docs-mcp Tool Usage Without Explicit <rules> in Claude Desktop

[YoungjaeDev]

Hi team,

I've configured langgraph-docs-mcp with Claude Desktop as per the provided documentation. The documentation states that I need to explicitly include in my prompt for LangGraph-related questions due to a lack of global rules support.

However, I've observed that Claude Desktop successfully utilizes the langgraph-docs-mcp server to answer LangGraph questions accurately, even when the block is not explicitly included in my prompts. For instance, questions like "How to create an agent in LangGraph?" are answered correctly without the explicit rules.

Could you please clarify why this is happening? Is there an undocumented change, an implicit mechanism at play, or am I misunderstanding the current setup?

Thanks.