Pre-check scan
How the scan works
Four layers, weighted scoring, grades — and what a static scan can and cannot claim.
The pre-check answers one question from the outside: if an AI agent arrived at your site right now, what would it find? It fetches the surfaces agents actually read — homepage, robots.txt, sitemaps, llms.txt, docs, OpenAPI, agents.md, your advertised MCP server — and runs 42 checks against them. No account, no code access, about ten seconds.
The four layers
Checks group into layers weighted 20/30/40/10. A layer with nothing applicable leaves the denominator entirely — a site is never punished for what it is not.
| layer | weight | the question it answers |
|---|---|---|
| Discovery | 20 | Can an agent find you and your documentation at all? Titles, robots.txt and agent-crawler policy, sitemaps, findable docs and API links. |
| Accessibility | 30 | Once an agent asks for a page, does it get usable content? Status parity for agent user-agents, readable text without JavaScript, llms.txt that parses and resolves, structured data. |
| Usability | 40 | Can an agent operate the product? A valid OpenAPI document with auth schemes and typed errors, machine-readable 404s, agents.md, and — where advertised — a working MCP server: handshake, listed tools, complete input schemas. |
| Payments | 10 | Can an agent transact? x402 / AP2 / ACP markers and 402 challenge behaviour. Most sites score n/a here — the honest answer; the layer can only raise a score. |
Scoring rules
- A passing check earns its full weight; a warning earns half; a failure earns nothing.
- N/A checks leave the denominator — they neither punish nor credit.
- Layer scores renormalise over the layers that had anything applicable.
- Grades: A+ ≥ 95 · A ≥ 86 · B ≥ 70 · C ≥ 48 · D ≥ 28 · F below. An unreachable site scores 0/F and says why.
- Every failing or warning check ships a concrete fix, written to stand alone.
Where scans live
Every scanned domain gets a permanent score page (/score/{domain}), a place on the leaderboard (reachable sites only), and an embeddable badge. Re-scanning updates the stored result; history is kept as score/grade points. The full per-check breakdown of the method lives on the methodology page, and the complete check catalog is the next page.