7 ms·
Not sure I understand what you mean. MCP servers and methods available do have to be visible to the model in some way so that it can spit out tool calls instruc
by JyB 2mo ago
Not sure I understand what you mean. MCP servers and methods available do have to be visible to the model in some way so that it can spit out tool calls instructions to be executed by the orchestration layer around it.
But you probably meant something else?
- wolttam 2mo agoMCP is the protocol by which the harness can connect with additional tools. The harness does the work of discovering the tools, the model is then fed a description of what tools it has available, with zero awareness of whether those tools are MCP or built into the harness itself (except for by the mcp__ prefix in the tool’s name, which a harness could choose to strip off)
- JyB 1mo agoRight. On a related note: "MCP is invisible to the LLM, or should be, under normal use". I am not convinced there are any downsides for the harness to explicitly feed MCP keywords into the LLM. It seem explicitly doing it has many advantages. For example, it allows the user to ask to interact with a resource through "the configured MCP" rather than directly through some local cli or HTTP endpoints; which can steer the LLM towards the proper actions. Otherwise, the model could in-theory be "clueless" with such a request.