[docs] Add Claude Code instructions to MCP docs

[Janpot]

Related page

https://mui.com/material-ui/getting-started/mcp/

Kind of issue

Missing information

Issue description

Add instructions for Claude Code. Additionally, for all clients, perhaps link to their relevant docs page about configuring MCP.

Context

No response

Search keywords: claude code

[Abhinav-Sriharsha]

Hi @bernardobelchior! I'm interested in contributing to this issue. Are you still working on adding the Claude Code instructions, or would you like me to take it on? Happy to collaborate if you prefer! Thanks

[bernardobelchior]

Hey, @Abhinav-Sriharsha. Feel free to take it on.

If this is your first contribution, take a look at CONTRIBUTING.MD as it might help with some doubts you might have

[saschabratton]

@Janpot @bernardobelchior @bharatkashyap

I have created PR #46621 which adds installation instructions for Claude Code.

However, the way the @mui/mcp tool is configured is not well suited for usage due to the structure of the llms.txt files.

See error from claude when trying to use the tool:

⏺ mui-mcp - useMuiDocs (MCP)(urlList: ["https://llms.mui.com/material-ui/7.2.0/llms.txt"])
  ⎿  Error: MCP tool "useMuiDocs" response (25712 tokens) exceeds maximum allowed tokens (25000). Please use pagination, filtering, or limit parameters to reduce the response size.

It would be better to break apart the sections into separate files, and summarize each section in the main llms.txt. It does not seem to be a public repo for the @mui/mcp module. Are you accepting contributions? If so, it is not clear how to do so.

[bharatkashyap]

@saschabratton This seems to be a Claude Code specific issue, which you can work around this with the MAX_MCP_OUTPUT_TOKENS setting https://docs.anthropic.com/en/docs/claude-code/settings#environment-variables

It would be better to break apart the sections into separate files, and summarize each section in the main llms.txt.

This is indeed how https://llms.mui.com/material-ui/7.2.0/llms.txt is currently structured, but there will be some changes soon around the structure of the file which will reduce the tokens further. Agree with your suggestion on open-sourcing the package, we will take that up internally.