Docs for agents

Point a coding agent at one of these instead of asking it to crawl the site. Every one of them is plain text served from the docs root, needs no auth, and is rebuilt whenever the documentation changes.

Which one

Wiring Shipeasy into a repo — /agents.md. It is the install → authenticate → keys → first flag → metrics → alerts path stitched in the order the work happens, followed by every CLI command and every MCP tool by name. That appendix is the part worth having in context: a hallucinated command name is the single most common way an agent fails at this.

Answering questions about Shipeasy — /llms.txt first, then fetch the one or two pages it points at. Cheaper and sharper than loading the corpus.

Bulk ingestion — /llms-full.txt, 92 pages in one request.

One page at a time

Every page on this site is also served as plain markdown at /md/<path>.md — /md/get-started/quickstart.md for the quickstart, 295 files in total, generated from the same conversion as the bundles above. The Copy page button under any page title copies exactly that file, and Open in hands the URL to an assistant.

Reach for it when you know which page you want. The corpus is for when you do not.

What is not in them

The 101 per-operation API reference pages and the 101 per-language SDK reference pages are indexed in llms.txt but not inlined — they are generated schema tables, and inlining them would bury the prose. Fetch the page you need, or read the OpenAPI spec directly.

Skip the fetch entirely

If your agent can run MCP, the shipeasy server ships docs_list, docs_get and docs_skill — the same content, retrieved by topic instead of by URL, plus installable skills that already know these workflows. See Install in your agent.