Skip to content

CONCEPT Cited by 1 source

Agent Readiness Score

Definition

Agent Readiness Score is a Lighthouse-style rubric proposed by Cloudflare on 2026-04-17 for grading a website on how well it cooperates with AI agents. The score is decomposed into four dimensions, each covering a small set of binary checks, plus a non-scoring Agentic Commerce layer:

  1. Agent Discovery — can an agent find the site's content? (robots.txt, sitemaps, Link: response headers per RFC 8288).
  2. Content for LLMs — is the content efficient for an LLM to consume? (markdown content negotiation default-checked; llms.txt opt-in).
  3. Access Rules — can the site differentiate friendly vs abusive bots and declare what AI can do with the content? (Web Bot Auth, Content Signals).
  4. Agent Actions — can an agent discover APIs, MCP servers, and skills the site exposes? (API Catalog (RFC 9727), MCP Server Card, Agent Skills index).

A fifth Agentic Commerce layer (x402, Universal Commerce Protocol, Agentic Commerce Protocol) is checked but does not count toward the score — explicit signal that agentic-commerce standards are emerging and intentionally unscored.

Canonical instance

The Cloudflare isitagentready.com scanner: enter a URL, Cloudflare makes live requests to detect which standards the site supports, returns a per-dimension pass/fail grid plus an aggregate score and actionable prompts the user can paste into a coding agent to implement each failing check.

Programmatic access: same check surfaces are available on Cloudflare URL Scanner as an Agent Readiness tab; same analysis via the URL Scanner API with options: {"agentReadiness": true}.

Why a score

Cloudflare's post frames the shape explicitly against precedent:

"Google Lighthouse scores websites on performance and security best practices, and guides site owners to adopt the latest web platform standards. We think something similar should exist to help site owners adopt best practices for agents."

The score exists to drive adoption of emerging standards — score-driven standard- adoption as a pattern, not just to provide diagnostics.

Current web-wide adoption (2026-04)

Measured by Cloudflare Radar on 200 k filtered top-visited domains:

Check Adoption
robots.txt present 78 %
Content Signals declared 4 %
Markdown content negotiation 3.9 %
MCP Server Cards + RFC 9727 API Catalogs combined < 15 sites in the entire dataset

Scoring notes

  • Weights not published. Cloudflare shows per-dimension pass/fail grids but doesn't disclose how the four dimensions combine to the headline score.
  • llms.txt is opt-in, not default-checked — because the default check (markdown content negotiation) covers the same "agent can get markdown at this URL" guarantee more directly.
  • Agentic commerce is non-scoring by design — the spec space is younger and the Agent Readiness Score deliberately avoids pressuring sites onto specific commerce standards while those converge.

Seen in

Last updated · 200 distilled / 1,178 read