Mux docs discovery

Mux publishes its entire documentation as agent-ready markdown, auto-generated from the same source as the real docs site. For any Mux question, fetch the relevant page and answer from it — never answer Mux API questions from memory, because API shapes and guidance change.

Workflow

1. Route — try these in order and stop at the first hit. The order is cost-ordered: each step reads fewer tokens than the next.
   1. If the mux CLI is installed, run mux docs find "<topic>" --agent — it searches the docs index CLI-side and returns only the matching page URLs, so you never read an index into context. (Check mux docs --help; older versions lack it.)
   2. Construct the URL from a known pattern: the tables below (collections, Robots robots-<task>.md, API references) cover most topics.
   3. Fetch the smallest collection index for the topic area (a few dozen lines).
   4. Fetch https://www.mux.com/llms.txt — the master index of every docs page — only when the steps above fail.
2. Fetch ONE page — the page's markdown version: append .md to its URL (e.g. https://www.mux.com/docs/guides/start-live-streaming.md). Only add an API reference bundle when you need exact request/response shapes; do not fetch extra pages "for context."
3. Answer from the fetched content and cite the page URL.

Never fetch https://www.mux.com/llms-full.txt. It is the entire documentation in a single file and will flood the context window; every page in it is reachable individually through the steps above.

Collection indexes