Agent Readiness Score
Your site is about to have visitors with no hands.
Search asked whether Google understood your page. AI search asked whether your answer got cited. The question now is whether an agent can finish the job — and most sites have never been measured on it.
One request, about 20 seconds. No sign-up, no email. The result gets a shareable address.
Understood
Read and indexed
Crawler permissions, content that survives without JavaScript, semantic structure and structured data — the difference between being summarised accurately and being summarised from your cookie banner.
Operated
Acted on
Forms with labels, controls with names, no bot challenge across the conversion path. An agent fills a form by reading it; a field called field_7 is a guess.
Integrated
Called as tools
WebMCP turns your actions into named, typed functions the visitor's own assistant can call — and a discovery document lets other systems find them before they arrive.
One number, immediately taken apart
A single score hides the thing you need to know. Ours splits in two the moment it appears — because “the site works but has adopted nothing” and “the tools are registered but the page is empty without JavaScript” are opposite problems that a single number reports identically.
| Axis | Points | Question | Pillars |
|---|---|---|---|
| Fitness | 65 | Does the site work for agents today? | Access · Readability · Actionability |
| Adoption | 35 | Which agent protocols are in place? | WebMCP · Interop & Commerce |
WebMCP shipped experimentally in Chrome in 2026 and adoption outside demos is still close to zero. A score that measured only WebMCP would hand almost every site on the web a nought, which tells nobody anything. So it carries 25 of 100 points — the largest single pillar, and the one that will decide the next two years — while the other 75 measure whether an agent can do anything with you in the meantime.
23 checks across five pillars
Every one is documented, every one shows its evidence, and checks that do not apply to your site are excluded rather than counted against you. Read the full methodology.
Agent Access
20 ptsBefore anything else can be judged, the request has to succeed. This pillar covers crawler permissions, bot management, consent walls and whether the page carries its content without a JavaScript engine.
4 checks
Machine Readability
20 ptsSemantic structure, structured data, content negotiation and the source files agents look for when they want a summary of your site instead of a rendering of it.
5 checks
Actionability
25 ptsThe heaviest pillar, because this is where revenue is. Forms an agent can fill, controls it can name, no bot challenge across the conversion path, and errors it can read.
5 checks
WebMCP Implementation
25 ptsThe core of the standard: named tools with typed schemas and honest annotations, registered against document.modelContext, so an agent stops guessing at your DOM and starts calling your functions.
5 checks
Agent Interop & Commerce
10 ptsDiscovery documents, agent-facing auth, bot identity and — for shops — the payment rails an agent needs to complete a purchase. Checks that do not apply to your site are excluded rather than counted against you.
4 checks
After the score
We fix what the check finds.
The report tells you what is wrong and gives you the code for most of it. When you would rather it was simply done — server rendering, form semantics, a first set of WebMCP tools, a discovery document that robots.txt does not hide — that is the work we do.