Can an AI agent actually use your website? We measured.
AgentReady runs real agents against real task flows — once through the raw UI, once through WebMCP tools — and scores every site 0–100. The gap between those two numbers is the case for the agent-native web.
Benchmark average today: 49/100. Sites exposing WebMCP tools average 100/100.
Every audited site, placed by readiness score. UI only exposes WebMCP our live demo store
| Site | Category | Score | Agent success (UI) | Agent success (WebMCP) | |
|---|---|---|---|---|---|
| AgentReady Outfitters (live demo store) | demo | 100 | 0% | 100% | detail |
| Harbor & Page Books sample data | books | 100 | 60% | 100% | detail |
| Tidepool Toys sample data | toys | 100 | 40% | 100% | detail |
| Kiln & Oak Home sample data | home-goods | 99 | 60% | 100% | detail |
| Loopset Fitness sample data | fitness | 55 | 60% | no tools | detail |
| Stitchport Apparel sample data | fashion | 52 | 60% | no tools | detail |
| Plumecraft Stationery sample data | stationery | 52 | 60% | no tools | detail |
| Saffron Basket Foods sample data | grocery | 48 | 60% | no tools | detail |
| Gridlamp Lighting sample data | home-goods | 36 | 40% | no tools | detail |
| Northline Electronics sample data | electronics | 22 | 20% | no tools | detail |
| Cartwheel Grocery sample data | grocery | 19 | 20% | no tools | detail |
| Monsoon Threads sample data | fashion | 19 | 20% | no tools | detail |
| Veltro Motor Parts sample data | auto-parts | 18 | 20% | no tools | detail |
| Brightfare Pharmacy sample data | pharmacy | 17 | 20% | no tools | detail |
| Fernway Travel sample data | travel | 0 | 0% | no tools | detail |
Rows marked sample data are illustrative fixtures shipped with the open-source repo so the methodology is inspectable — they are not measurements of real companies. The demo store rows are live, reproducible runs.
Readiness report
Ask your agent to draft a report here (draft_readiness_report). It lands in this editor; you refine it; the agent can read your edits back (get_report_draft) and revise. Same document, two authors.
Agent activity on this page
Every WebMCP tool call is logged by the AgentReady SDK — this is the analytics layer working on its own dashboard.