SI Readiness by MCP-Tool
How the score is calculated
Fitness measures current usability; adoption measures published agent interfaces. Checks that do not apply are excluded rather than counted as failures. The assessment reads public resources and does not sign in, submit forms or execute a purchase.
Agent Access · 20 points
- robots.txt lets AI crawlers in · 5 points
Does robots.txt allow the crawlers behind AI answers to read the page?
- Agent user agents are actually served · 5 points
Does the server answer when the request carries an agent user agent?
- No wall in front of the content · 4 points
Does the content arrive without a consent gate or bot challenge in the way?
- Content survives without JavaScript · 6 points
Is there real text and a heading in the raw HTML, before any script runs?
Machine Readability · 20 points
- Semantic landmarks and heading order · 4 points
Are there landmarks and a heading hierarchy that does not skip levels?
- Valid structured data · 6 points
Is there parseable JSON-LD, and does its type match what the page is?
- Markdown content negotiation · 4 points
Does the page return Markdown when a client asks for it?
- llms.txt source file · 3 points
Is there a curated file that tells an agent what this site is and where to look?
- Sitemap and canonical URL · 3 points
Is there a valid sitemap, and does the page declare which address is the real one?
Actionability · 25 points
- A primary action is discoverable · 5 points
Is there something on this page an agent could complete on the visitor's behalf?
- Forms an agent can fill · 7 points
Do the fields carry labels, names, types and autocomplete tokens?
- Controls have accessible names · 5 points
Can every interactive control be identified by name rather than by position?
- No bot challenge on the conversion path · 5 points
Is the main action free of CAPTCHA or a bot challenge?
- Errors an agent can read · 3 points
Are validation states expressed in markup rather than only in colour?
Integration · 15 points
- An MCP endpoint an agent can reach · 7 points
Can an agent call your capabilities without a browser being open?
- A machine-readable description of your API · 5 points
Can an agent learn your API without a human reading your docs?
- An agent card other agents can read · 3 points
Can another agent discover what you are and what you can do?
Identity & Commerce · 10 points
- Agent-facing authentication · 4 points
Is there a documented way for an agent to authenticate?
- Agentic payment rails · 3 points
Can an agent complete a purchase without a human at the keyboard?
- Bot identity and content signals · 3 points
Can you tell a legitimate agent from an impostor, and have you stated your terms?
WebMCP Implementation · 10 points
- document.modelContext is in use · 6 points
Does the page register WebMCP tools at all?
- Declarative tools on forms · 4 points
Are existing forms annotated so the browser can synthesise tools from them?
- Tool names and descriptions · 5 points
Are the tools named legally and described well enough to choose between?
- Typed input schemas · 5 points
Does each tool declare a typed JSON Schema with described properties?
- Honest annotations on tools · 5 points
Are read-only tools marked read-only, and is anything destructive gated behind confirmation?