# AgentCapable

> Free agent-readiness scanner. It measures how a website's edge treats AI
> agents: 14 tracked agent User-Agents probed from two independent datacenter
> networks plus three controls, judged against what the site's robots.txt
> declares, and 23 further checks that grade how readable the site is for
> agents. Deterministic rubric — no AI in the score. Every scan is fresh;
> results are never served stale.

## Using the API (no key; contact details required)

POST https://agentcapable.dev/scan with JSON
{"domain": "example.com", "lead": {"name": "...", "email": "...", "phone": "...", "role": "optional"}}
returns the full scan report as JSON: headline verdict (open |
partially-blocked | agent-hostile | contradictory | unknown), reasons, a
per-agent access matrix (robots.txt decision vs per-network edge outcome per
User-Agent), the 23 scored checks with evidence, a 0-100 readability score
(or "?" when there is not enough signal), and the probe method statement.
Add {"fresh": true} to bypass the 45-second double-submit window. Rate
limit: a few scans per domain, then about one per minute.

## Access

Every scan — site or API — requires contact details (name, email, phone;
role optional), submitted in the request body as shown above and stored with
the scanned domain. Requests without them receive a 400 naming the missing
field.

## Pages

- [Scan a domain interactively](https://agentcapable.dev/)
- [This site as markdown](https://agentcapable.dev/index.md)
- [Scanner disclosure, politeness policy, abuse contact](https://agentcapable.dev/bot)

## Honesty rules

Findings are capped at what datacenter-network evidence supports: the
strongest claim is "this edge treats agent-identifying requests from cloud
servers as hostile" — never that a vendor's verified bot is blocked. A site
that declares a block in robots.txt and enforces it consistently gets full
credit. A site we cannot measure is "unknown", never a low score.