Why do I still need to configure an LLM model when using Claude Code to connect to an MCP server?
In this mode, shouldn’t MCP only be used as a retrieval/tool-calling capability? In theory, it should just provide access to external data or tools, rather than requiring a separate LLM model configuration.
Why do I still need to configure an LLM model when using Claude Code to connect to an MCP server?
In this mode, shouldn’t MCP only be used as a retrieval/tool-calling capability? In theory, it should just provide access to external data or tools, rather than requiring a separate LLM model configuration.