Open navigation
Is your site ready for agents?
Find out in seconds.
One address in, a score out, and a concrete fix for every gap. Works for any site, whatever it's built with.
Ten checks.
One score.
Every check looks at what an agent sees without a browser: the HTML, the Markdown, the discovery files, and the tools a site offers. Scores add up to 100; drafts are reported but not scored.
Can agents get in?
35 pts- 20 pts
Readable without JavaScript
Content in the HTML, not behind a sign-in or a client render.
GET / - 15 pts
robots.txt
Agents acting for users and AI search aren't blocked.
GET /robots.txt
Can agents read it?
50 pts- 20 pts
Markdown version
A .md URL and Accept: text/markdown, linked from the page.
GET /index.md - 15 pts
llms.txt
A Markdown index of the site for language models.
GET /llms.txt - 10 pts
Title, canonical, Open Graph
How agents name and cite the page.
<head> - 5 pts
JSON-LD
schema.org data about who runs the site.
ld+json
Can agents use it?
15 pts- 10 pts
MCP server
Tools agents can call over Streamable HTTP.
POST /api/mcp - 5 pts
OpenAPI
A machine-readable description of the site's API.
GET /openapi.json - Reported
Agent card
Reported, not scored, while the specs settle.
/.well-known - Reported
WebMCP
Tools the page registers for in-browser agents. Reported.
modelContext
Readable is step one.
A new Farm.js app passes the reach and read checks out of the box: Markdown, llms.txt, JSON-LD and Open Graph come from its pages and metadata, and MCP and OpenAPI are a config block away. The agent infrastructure we're building goes further: one gateway for your tools, and a trace of every agent run.