AgentReady — a 24-point audit of whether a site is readable by agents (free to run)

Built this today because the agent-readability problem is boring and fixable and nobody measures it.

AgentReady fetches a URL server-side and scores 24 signals across five groups: agent access (robots.txt, AI crawler allowlist, llms.txt, sitemap, X-Robots-Tag), machine-readable structure (JSON-LD, OpenGraph, meta, canonical, lang, feed), extractability (text-to-markup ratio, server-rendered content, single H1, landmarks, alt coverage, internal links), freshness/trust (dates, author, HTTPS), and performance (TTFB, weight, viewport).

Weighted 0-100, graded A-F, every failure ships with a concrete fix plus a ready-to-paste llms.txt and robots.txt patch.

Live: https://isagentready.vercel.app Source: https://github.com/zachperk/agentready API: POST https://isagentready.vercel.app/api/audit {"url":"example.com"}

Sample scores: stripe.com 76, github.com 69, example.com 35, itself 90/A.

Free view is the score plus the three worst issues. Full report is $10, honor-system unlock, no account or email. If you run it on your own site and the score is bad, the fixes are all in the report — that is the whole product.

Feedback welcome, especially on the weighting: right now blocking GPTBot costs 10 points and client-only rendering costs 10, which is roughly the right ratio but I am not certain.