docs-navigation

To navigate hierarchical documentation indexes efficiently, start at the root index, match task keywords against index files, and drill down to the exact page needed. This protocol-driven flow ensures deterministic results while loading only necessary content to preserve token budget.

Local-first documentation navigation is the process of reading local index files to locate relevant pages before performing a web search. You need this approach when working with tracked libraries like Claude Code, BAML, or MCP to minimize context loading and reduce token usage.

To search local ai-docs indexes using keywords, read the _index.toon files within the hierarchy, match your task keywords against the indexed terms, and load only the specific pages that surface as relevant. This method avoids guessing paths and prevents loading full-context content unnecessarily.

Yes, this navigation approach supports tracked resources including libraries like Claude Code, BAML, and MCP. It operates within the ai-docs hierarchy by reading index files and drilling down to precise pages, ensuring accurate and efficient results across these tracked documentation sets.

The best way to minimize token usage when searching documentation indexes is to enforce a strict navigation protocol that avoids guessing paths. By starting at the appropriate index level and loading only the necessary pages rather than full-context content, you maximize efficiency and conserve your token budget.

Documentation navigation fails when paths are guessed because the ai-docs hierarchy requires reading _index.toon files to locate relevant pages accurately. Guessing paths bypasses the protocol-driven flow, leading to non-deterministic results, excessive token consumption, and failure to load the necessary content.