AIDOC
Token-optimized site maps for AI agents. A compact JSON map that cuts 85–95% of the tokens a sitemap would cost.
Agents pay to understand a site they could have just been told about
An XML sitemap lists URLs. It says nothing about what a page is for, what an agent can do there, or where it should go next. So agents crawl HTML, burn tokens reconstructing structure a site already knows, and still guess. The map exists. It's just written for crawlers from 2005, not for models paying per token.
A map written for the reader
AIDOC is a single compact JSON document served at a well-known path, /site.aidoc. Single-letter keys and intent-first fields describe each page's purpose, available actions, and navigation, so an agent reads the whole site in one cheap request instead of crawling it. In practice that's an 85–95% token reduction versus a traditional sitemap.
{"v":"1.0","d":"hariprasd.me","t":1737374400,
"p":[{"u":"/","f":"home","a":["nav"],"n":["/works","/contact"]},
{"u":"/works","f":"work index","a":["nav"]}],
"m":{"author":"Hariprasad B","standard":"AIDOC 1.0"}}
Every key earns its place: v version, d domain, t timestamp, p pages, m meta. Per page: u url, f function, a actions, n navigation, d dynamic. Nothing an agent doesn't need.
This site eats its own dog food
The portfolio you're reading ships a live, build-time AIDOC generated from its own content. Every work in this index is already in the map. Fetch it:
GET https://hariprasd.me/site.aidoc
If a standard is worth proposing, it's worth running in production. So this one does.