# Agents

Operating notes for AI agents working with https://terrabillion.com. Read llms.txt first for when-to-use guidance.

## Reading pages

- Append nothing: request any page with `Accept: text/markdown` for a Markdown rendering (same URL, `Vary: Accept`).
- Missing pages answer HTTP 404 in Markdown with recovery links. Never treat a 404 body as content.
- Feeds: /blog/rss.xml lists every post. /sitemap.xml lists every crawlable page with lastmod dates.

## Discovery files

- /llms.txt: when-to-use guidance and the doc index.
- /llms-full.txt: the guide plus every post.
- /.well-known/ard.json: ARD manifest of agent-facing artifacts (legacy alias /.well-known/ai-catalog.json).

## Authenticated surfaces

- The bid queue, library records, and JSON APIs need a signed-in subscriber (email magic link or Google). There is no API key or agent credential flow; agents act through the signed-in user's browser session.
- Rate limits answer HTTP 429 with Retry-After. Back off and retry; do not parallelize around it.

## Contact

- Product questions and access requests go through /contact. No other inbox is monitored for agent traffic.