WebMCP explained: should your website expose tools to AI agents?
Instead of an agent guessing which button to click, your page tells it: "here is a search tool, here is a booking tool, here are the inputs". It is promising, experimental and changing monthly.
Today an AI agent uses your site the way a careful stranger would: read the page, find the button, hope it does what it looks like. WebMCP proposes a contract instead.
What WebMCP is
Chrome describes WebMCP as a proposed web standard to help sites build and expose structured tools for AI agents. A tool has a name, a description, an input schema and a function that runs on the page, visibly, so the user can see what the agent does.
| API | How you define a tool | Good for |
|---|---|---|
| Imperative | Register a tool with JavaScript | Search, filtering, multi-step actions |
| Declarative | Annotate an existing HTML form | Sign-up, contact, booking forms |
Status in October 2026
- Origin trial: Chrome lets sites join from Chrome 149. Reporting puts the trial window at Chrome 149 to 156.
- Local testing: enable
chrome://flags/#enable-webmcp-testingand relaunch Chrome. - Permissions: a
toolsPermissions Policy defaults toself; cross-origin iframes needallow="tools". - Lighthouse: the Agentic Browsing category includes WebMCP audits (registered tools, forms missing declarative WebMCP, schema validity), which need the origin trial to run.
- Standard status: a community group draft, explicitly not on the W3C standards track. The API has already been renamed once, so expect changes.
Chrome's own page lists the limits: WebMCP is designed mainly for local browser workflows with a human in the loop, and an agent has to visit your site to discover its tools.
Should you implement it?
| Your situation | Recommendation |
|---|---|
| Forms lack labels or rely on scripts | Fix those first; declarative WebMCP builds on well-labelled forms |
| Complex search, configurators or booking | Prototype one imperative tool behind the origin trial |
| Simple brochure site | Wait; good HTML already does the job |
| Already have an API or MCP server | Keep it; WebMCP is for in-browser agents, not a replacement |
How to prepare without the trial
- Make every form field and button readable to agents. Check with our agent form checker.
- Write down your site's three core tasks as if they were tools: name, inputs, output.
- Keep task flows simple: no disabled buttons waiting on scripts, clear validation messages.
- Run Lighthouse's Agentic Browsing category in Chrome 150 or later to see where you stand.
For the bigger picture, see what AAO is.
Questions people ask
What is WebMCP?
A proposed web standard that lets websites register structured tools (name, description, inputs, function) that browser AI agents can call directly, either from JavaScript or by annotating HTML forms.
Is WebMCP available in Chrome?
As an origin trial from Chrome 149, and locally behind the chrome://flags/#enable-webmcp-testing flag. It is experimental and subject to change.
Do I need WebMCP to work with AI agents?
No. Agents can use well-built HTML today. WebMCP may make complex tasks more reliable, but readable, well-labelled pages come first.
Want this mapped onto your product?
See how our AAO work runs, or send your store link and a founder will reply personally.
