Could the server describe the tool's output schema in tools/list?

[gsabran]

Pre-submission Checklist

• I have verified this would not be more appropriate as a feature request in a specific repository
• I have searched existing discussions to avoid duplicates

Your Idea

Reading the specs, it seems that the server only specifies the schema for the input it is expecting a tool to be called with. It doesn't specifies the schema for its response.

I was curious about the rational behind limiting the schema to only the input. Most API schemas, including JSON based one (ie GraphQL), describe both. Doing so is particularly useful for the client to help ensure that it can handle all types of response. The alternative seems to be making a few requests, looking at the different shapes of response that come back, and building support for those. This is brittle as other shapes of response can be returned in production and the client would then break.

I'm sure adding a schema for the response has been evaluated when writing the specs. Before suggesting that MCP should require this to be done, I was curious to learn about why this was not part of initial version.

Scope

• Protocol Specification
• SDK Features
• Documentation
• Developer Experience
• Other

[gsabran]

I now see that only a few specific response types are expected from the server.