Grok MCP

Implementation Reference

• src/server.py:480-500 (handler)

  The implementation of the `stateful_chat` MCP tool, which initializes an xAI client, configures chat parameters, and samples a response.

  @mcp.tool()
  async def stateful_chat(
      prompt: str,
      model: str = "grok-4-1-fast-reasoning",
      response_id: Optional[str] = None,
      system_prompt: Optional[str] = None
  ):
      client = Client(api_key=XAI_API_KEY)
      
      chat_params = {"model": model, "store_messages": True}
      if response_id:
          chat_params["previous_response_id"] = response_id
      
      chat = client.chat.create(**chat_params)
      if system_prompt and not response_id:
          chat.append(system(system_prompt))
      chat.append(user(prompt))
      
      response = chat.sample()
      client.close()