OptionalannotationsHints an agent may use to decide how carefully to call this tool: readOnlyHint,
destructiveHint, idempotentHint, openWorldHint, title.
They are hints, never a guarantee, and never a substitute for the authorization on the route.
OptionaldescriptionWhat the tool does, in the words a model reads before deciding to call it.
Derived from the route's openapi summary or description when it is not stated here. A tool
with neither is still exposed, and says so in the log: an agent given a bare name will guess,
and guessing against a route that writes something is the failure worth seeing early.
OptionalinputThe arguments the tool takes, as JSON Schema.
Derived from the route's validation schema when it is not stated here, and from the route's path parameters when there is no schema to read. State it to override both.
OptionalnameThe name an agent calls. Defaults to the route's name.
OptionaloutputThe shape of the result, as JSON Schema.
When declared, a successful call also answers structuredContent, so an agent reads a value
rather than parsing prose. Undeclared, the result travels as text only, which is what the
protocol asks for when a shape was never promised.
The long form of a tool declaration.