Audit your site's WebMCP

We load your site in a real browser and score the tools it gives an agent. Below, every audited site ranked by score; the full directory lists the rest: community registry claims, and sites our audits have observed live.

Agent goal

#	Site	Score	Tools	Tool names
1	Sentrysentry.io	99	1	start_signuprequest_demo+4
2	Monday.commonday.com	97	1	navigate_by_use_case
3	navattic.com	95	2	navattic_open_marketing_urlnavattic_scroll_to_top
4	Renderrender.com	91	5	render.docs.searchrender.docs.get-markdown+3
5	reebok.com	84	11	search_catalogbrowse_store+8
6	feverup.com	84	5	toggle_plan_favoritehome_list_cities+2
7	hostinger.com	84	2	list_hostinger_pagesnavigate_to_hostinger_page
8	Forterforter.com	81	6	list_skillsget_openapi_spec+4
9	Razorpayrazorpay.com	80	1	get_product_details
10	speedof.me	79	4	run_speed_testcreate_share_link+2
11	ziprecruiter.com	79	1	landing.search
12	Telnyxtelnyx.com	77	4	ask_telnyxget_pricing+2
13	aloyoga.com	76	11	search_catalogbrowse_store+8
14	awaytravel.com	76	11	search_catalogbrowse_store+8
15	chainstack.com	74	6	chainstack.get_site_infochainstack.search_content+4
16	tatum.io	74	3	discoverTatumMcpServeropenTatumProduct+1
17	marketresearchfuture.com	72	4	search_reportsview_report+2
18	chilipiper.com	72	3	search_chilipiper_siteget_page_as_markdown+1
19	Customer.iocustomer.io	67	3	open_customerio_pagestart_customerio_free_trial+1
20	usercentrics.com	66	4	uc_read_pageuc_find_regulation+2

Browse the full directory (1,422 sites)

How the score works

Every audit is scored on four pillars: Shared experience, Tool selection, Tool quality, Trust, out of 100.

Shared experienceweight 30

People and agents share this page. We grade the page a person actually sees and whether they can see what the agent does, so a site cannot score well by serving agents and no one else.

Tool selectionweight 25

Can an agent pick the right tool here. We give an agent the tasks a site of this kind exists for and check it selects a registered tool with valid arguments, and that the tool set covers the job. Nothing is executed yet, so this measures selection, not completion.

Tool qualityweight 25

Are the tool contracts built right. Valid schemas, descriptions that say what comes back, names an agent can tell apart, handlers behind every tool.

Trustweight 20

Can an agent trust what the tools declare. Read-only hints that match what the tool actually says it does, and metadata that describes tools instead of steering the agent reading it.

The open source webmcp plugin teaches your coding agent to audit a site, implement tools on document.modelContext, and verify them in a real browser. npx @ora-ai/webmcp-verify runs the verification on its own. No signup, no hosted service.Entering the OpenAI WebMCP Challenge? An audit here checks, in a real browser, whether an agent can find and call your tools: page experience, tool contracts, tool selection. It does not execute them, so end-to-end execution and originality still take human review.

Audit before you publish

npx ax webmcp-audit http://localhost:3000

The ax CLI runs this same audit against a page on your machine. Your browser captures the tools, ora scores them, and nothing is stored or ranked.