Skip to main content
Everything PalatineAI publishes is static files over HTTPS. No auth, no API keys, no pagination, no enforced rate limits — be reasonable and cache.

​
Machine-readable surfaces

SurfaceURLContent type
Datasethttps://palatineai.site/projects.jsonapplication/json, ~140 KB
Agent skillhttps://palatineai.site/skill.mdtext/markdown, ~4 KB
Site map for LLMshttps://palatineai.site/llms.txttext/plain — links every published artifact
This documentationany docs page + .md suffixraw markdown of the page
Docs index for LLMs/llms.txt and /llms-full.txt on this docs siteall pages, linked or inlined
Every docs page also has a contextual menu (top right) — copy as markdown, or open the page directly in Claude, ChatGPT, or Cursor.

​
Freshness without re-downloading

Both files serve ETag and Last-Modified. Poll cheaply with a conditional request:
Conditional fetch — 304 if unchanged
curl -fsS https://palatineai.site/projects.json \
  -H 'If-None-Match: "6a2abceb-22cfd"' \
  -o projects.json -w '%{http_code}\n'
Store the ETag from each 200 response and send it back as If-None-Match; a 304 costs almost nothing. Inside the JSON, stats.updated (YYYY-MM-DD) is the dataset’s own version stamp — pin it in any conclusions you log.

​
Transport details, honestly

  • CORS: projects.json does not currently send Access-Control-Allow-Origin. Fetch server-side (agents, CLIs, backends — all fine); direct browser fetch() from another origin will be blocked.
  • Stability: URLs above are the contract. Fields are additive — new keys may appear in projects.json; parse leniently and don’t fail on unknown keys.
  • Size: the whole feed is one ~140 KB GET. Don’t scrape the HTML dashboard — it renders from the same projects.json you can fetch directly.