Agents Are Coming to the Web.Is Your Product Ready?
Scan your site. See what agents can — and can’t — do with it today.
WebMCP is a W3C web standard backed by Google, Microsoft, and the Web Machine Learning Community Group. It gives AI agents a structured, deterministic interface to interact with your product. No screen-scraping, no guesswork.
Try WebMCP live*
Every form on this page is a WebMCP tool. Click “Try it live” to watch an AI agent discover and execute each tool in real time.
scan_webmcp_readinessScan any website URL and return a structured WebMCP readiness report with score, grade, and actionable findings.
Parameters
url:stringreqjoin_waitlistJoin the WebMCPReady early-access waitlist. Submits the visitor email and any scan session ids collected from readiness scans.
Parameters
email:stringreqbook_discovery_callBook a WebMCP discovery call with invitee details pre-filled. The visitor still chooses a meeting time.
Parameters
firstName:stringreqlastName:stringreqemail:stringreqcompanyWebsite:stringreqrole:stringrequrgency:stringreqcurrentStack:string* This is a simulated demo showing what AI assistants — like Cursor, Claude Code, or ChatGPT — could do on this site using WebMCP tools.
Agents use your product like a confused first-time visitor.
WebMCP gives them expert-level proficiency.
Products that expose their core flows to agents gain a native advantage. Agents can discover, recommend, and execute actions in your app without workarounds. Without it, agents guess at your UI — retrying, failing, and skipping flows entirely.
W3C Proposal · Early Preview in Chrome 146We’re building a tool that turns your key user flows into agent-ready capabilities. Drop your email. We’ll let you know when it’s ready.
Why it matters
WebMCP replaces guessing at UI elements with structured tool calls. Agents execute your flows deterministically. No vision model, no fragile selectors, no retries.
Expose flows with HTML attributes on existing forms or a few lines of JavaScript. No backend changes, no new services. It runs in the user’s browser.
Products that agents can use natively get recommended, automated, and integrated first. Those that can’t become invisible to the agentic workflow.
WebMCP is a W3C-incubated standard with an early preview in Chrome. Build on an open spec that works across agent platforms. No vendor lock-in.
What we measure
A comprehensive check across three levels: from basic reachability to full WebMCP tool integration.
Can an AI agent even reach your page?
- Secure HTTPS connection
- Page loads with valid HTTP status
- Correct HTML content headers
- Not blocked by
robots.txtorX-Robots-Tag - Fast response, minimal redirects
The baseline. If agents can't even load your page, you're invisible to the entire agentic workflow.
Can AI agents find and understand your site?
robots.txtpresent & permissive- Valid web app manifest
llms.txtfor LLM contextai.txtfor agent policies
Agents need to know what your product does before they can recommend it. Discovery metadata is how they find you.
The structured interface AI agents actually call.
- Declarative form tools via HTML attributes
- Imperative JavaScript API (
navigator.modelContext) - Feature detection and graceful fallbacks
This is where your product goes from 'visible' to 'usable.' Agents call structured tools instead of guessing at your UI.
We’ll review your web app, assess its agent-readiness, and identify the most valuable user flows specific to your product to expose to AI agents.
Frequently asked questions
WebMCP Agent
Live demo