Can software act on your website, or only read it?
A draft specification for letting AI agents act on websites, WebMCP, is being written in public. Whether assistants will book services this way, and how soon, is not yet known.
Below is how it works, for the technically curious.
The 3-Layer AEO Stack
Layer 1: Crawlability
Foundationrobots.txt, sitemap, meta tags
Search engines can find and read your website.
Layer 2: Understanding
Where Bodyne works todaySchema, AEO, llms.txt
Search engines and AI platforms understand your products, services, and expertise through structured data.
Layer 3: Action
Experimental: draft specificationWebMCP tools, llms.txt
AI agents that support WebMCP can see what your site does and take action on behalf of users.
What WebMCP Is
WebMCP is a draft Community Group Report from the W3C Web Machine Learning Community Group, edited by engineers at Google and Microsoft (spec draft). It is not a W3C Standard and is not on the W3C Standards Track. It describes a way for a web page to expose callable tools to AI agents.
Pages register those tools in JavaScript with document.modelContext.registerTool() on a secure (HTTPS) page. In Chrome, support is an origin trial (Chrome developer blog). The spec does not define a manifest file or a /.well-known/ path. Instead of listing pages for crawlers, tools describe actions that agents can take.
Websites register "tools" that agents can call: request a quote, search products, check compliance status, schedule a consultation.
How It Works
Business registers WebMCP tools on its pages (for example on the quote or emergency form)
Registered tools describe what an agent can do (e.g., "request_quote", "search_products", "book_consultation")
A browser-based AI agent visiting the page sees the tools
User asks agent: "Find me a restoration company near me" or "Who makes fire extinguisher racks?"
Agent invokes the tool, gets structured response
Agent recommends the business — with ability to book or request a quote directly
Why This Matters for Your Business
A homeowner may one day tell an AI assistant to handle the flood. Whether the assistant can act on your business will depend on what your site exposes. That is the work.
Frequently Asked Questions
What is WebMCP?
WebMCP (Web Model Context Protocol) is a draft specification from the W3C Web Machine Learning Community Group (not a W3C Standard) that lets web pages expose callable tools to AI agents. Instead of listing pages, it describes things agents can do on your site.
Do I need WebMCP if I already have good SEO?
Good SEO (Layer 1-2) makes you findable. WebMCP (Layer 3) makes you actionable. An AI agent can find your site with good SEO. WebMCP is designed so that, where an agent supports it, the agent can request a quote, search your products, or check compliance on the user's behalf.
How much does it cost to implement?
The effort depends on which tools you want to expose. Agent Readiness work — tool design, scoping, implementation — is included in The Enforcement ($15,000, 90 days) alongside the rest of the 6-surface build.
Is this just for large companies?
No. The draft specification does not depend on company size or industry. Any page served over HTTPS can register tools.
When will AI agents actually start using this?
Not at scale yet. WebMCP is a draft, and Chrome support is currently an origin trial. Engineers at Google and Microsoft edit the spec. Nobody can say how fast it will spread, so we build on what already works today (structured data and llms.txt) and add WebMCP tools where a client has a form worth exposing.