AI Developer Toolkit documentation

search and fetch are cleanly separated: search returns up to 10 ranked snippets with ids, fetch retrieves the full markdown for exactly one id. The descriptions explicitly cross-reference each other (fetch requires an id from search), so an agent cannot reasonably misselect between them.

Both tools use a single lowercase imperative verb with no mixed casing or alternate verb styles. The two names are short, parallel, and follow the same predictable convention throughout the set.

Two tools is below the typical 3-15 sweet spot, but for a read-only documentation corpus the search+fetch pair is the minimal complete capability set and each tool clearly earns its place. No extraneous tools are present.

The retrieve-lifecycle is covered: discovery via search and full-text retrieval via fetch, which is the full CRUD-equivalent surface for a read-only docs corpus. Minor gaps exist around result pagination, browsing the index, or fetching multiple articles at once, but agents can work around these with narrower queries and repeated fetch calls.