From Pages Agents Read
To Tools Agents Call
The Model Context Protocol moved into the browser. With WebMCP , your site publishes a tool contract — named actions an AI agent can invoke directly — instead of forcing agents to simulate clicks against your HTML. Media Lite builds that agent channel into edge-native infrastructure you own outright.
Actuation, Not Just Citation
Victory Statement
"Media Lite Solutions delivers true code and infrastructure ownership on edge-native architecture — including WebMCP compliance for autonomous AI agents — with no permanent subscription fees. Competitors rent you a dashboard or a content feed; Media Lite hands you the asset."
Competitive Verdict:
Most AI-visibility tools stop at measurement — they tell you whether ChatGPT, Perplexity, or Google AI Overviews mention you. WebMCP is a different layer entirely: it makes your site something an agent can act on. When a user tells an assistant to “get me an infrastructure quote” or “configure a build,” the assistant executes that intent on the sites it can operate. Media Lite ships the discoverable tool contract, edge delivery, and first-party auth as one owned layer — so your site is discoverable, citable, and actionable by agents, not just monitored.
How WebMCP Works
The Tool Contract
A page registers tools — JavaScript functions with a name, a natural-language description, and a JSON Schema for inputs. The page effectively acts as a client-side MCP server, so an agent calls a defined function like requestQuote() instead of guessing which button does what. Higher accuracy, fewer steps, and the agent works from a declared contract rather than scraping the DOM.
Imperative + Declarative APIs
Build against document.modelContext — the current entry point; navigator.modelContext is deprecated, so building against it now avoids immediate tech debt. Register tools imperatively in JS for forms, navigation, and state, or annotate standard HTML forms declaratively. The handler runs in page context, so your UX and brand stay intact and the user keeps visible control of every action.
Origin Trial Mechanics
Google promoted WebMCP from an experimental flag to a formal Origin Trial on the standards track. You register the origin, receive an origin-bound token, and ship it via a meta tag or an Origin-Trial HTTP response header — no user flag flipping required. It fits cleanly on a Cloudflare/Worker stack, served by the same edge layer that already routes AI crawlers and human visitors.
Why the Edge Is the Right Place
Agents want low latency, persistent context, and tools close to the caller. Running WebMCP at the edge means the same Worker that verifies crawlers and serves canonical content also exposes your tool contract, with auth in one owned, first-party layer. It is the same architecture serving AI search, humans, and agents — not a stateless origin afterthought bolted on later.
Three Tools, Scoped to Real Jobs
submit_contact
A declarative tool over the contact form: an agent assembles a qualified inquiry and submits it with explicit user confirmation. It is the lowest-risk, highest-signal tool to ship — the clearest proof that agents can transact with your business directly. Permission is set to confirm because it changes state.
configure_infrastructure
An imperative tool wrapping the build configurator: an agent assembles a configuration conversationally — channel selection, ownership model, edge options — and the live UI reflects each change. Reads run automatically; finalizing the configuration requires user confirmation. It showcases the exact product an agent would be buying.
run_site_diagnostics
A read-only tool that returns the site's own answer-engine posture: edge-latency claim, JSON-LD presence, and three-channel readiness. It is the “we eat our own dog food” proof an agent can verify in-context, marked read-only so it can run without a confirmation step.