About AgentPass
A small, opinionated index of whether software can read the web the way people do. Built and maintained by Scott Scheferman.
Why this exists
Agents fail on the web for boring reasons: a robots group that hides the catalogue, a product page that needs JavaScript to state a price, a canonical tag pointing at a staging host, a contact page that only exists inside a chatbot. None of these are hard to fix. All of them are hard to notice, because a browser shows a person a perfectly good page while an agent receives an empty mount node.
AgentPass narrows that gap to a repeatable measurement. It fetches what a crawler fetches, scores what came back against a fixed rubric, and writes the fix as a prompt an agent can act on. No browser rendering, no login, no model in the scoring path.
How it is built
A single Node process with no runtime dependencies. The engine is a plain HTTP fetcher with a deterministic scoring table; the site you are reading is server-rendered HTML, and the same reports are available as JSON, Markdown and over MCP. The registry of checks, crawlers and weights is public and versioned with the engine.
What it will not tell you
It will not tell you whether an agent understood your page, whether a language model will quote you, or whether your traffic will rise. It measures affordances: the artefacts an agent needs before comprehension is even possible. Treat a clean score as a floor, not a forecast.
Contact
Email hello@agentpass.dev for scan disputes, false positives, or a crawler we have missed. Include the host and the report URL so the evidence line can be checked against a fresh fetch. We answer within one working day, and every disputed check gets a dated entry in the changelog.
For machine consumption: llms.txt, AGENTS.md, OpenAPI 3.1, and the MCP endpoint at /mcp.
Standards and sources
- RFC 9309 — Robots Exclusion Protocol
- RFC 9727 — API discovery linkset
- RFC 9728 — OAuth protected resource metadata
- llms.txt convention
- Schema.org vocabulary
Run a scan
One URL, free, no account. The crawler matrix alone is usually worth the trip.