A server author wants to give the language model natural-language guidance on how to use the server as a whole, not tied to any single tool. Where does revision 2026-07-28 provide for this?
Select an answer to reveal the explanation.
Short Explanation
Server-wide advice belongs on the cover page, not copied into every chapter. In 2026-07-28 that cover page is the instructions field in server/discover. If you find yourself pasting the same paragraph into every tool description, you've picked the wrong spot.
Full Explanation
The DiscoverResult returned by server/discover contains supportedVersions, capabilities, the server's identity in _meta, and an optional instructions field described as natural-language guidance for LLMs on how to use the server effectively. That is the place for server-wide advice, such as which tools to call first or how the server's resources relate to each other; a host can include it in model context when the server is in use. Creating a prompt named instructions misuses prompts: they are user-controlled templates that a person chooses to invoke, and nothing obliges a host to run a particular prompt before each conversation. Repeating server-wide guidance in every tool description bloats each tool definition and mixes global advice with what a single tool does; descriptions should explain the individual tool. The serverInfo name is a short identifier that is self-reported and intended for display, logging and debugging, and clients should not change behaviour based on it. Exam caveat: hosts decide whether and how to surface instructions to the model, so the field guides behaviour rather than guaranteeing it. Operational check: call server/discover, confirm the instructions text is present and concise, and verify that the host actually places it in the model's context when the server is attached.