4
Overview

How AnyWebMCP works

AnyWebMCP is a Chromium extension that adds structured, site-specific tools to supported websites. A WebMCP-capable agent can discover and call those tools inside the current browser session. The agent works with focused results and avoids scanning the page's full HTML.

Architecture

From website to agent tool

Supported website Site package WebMCP Your agent

  1. The extension matches the website

    When you open a supported domain, the extension loads the integration packaged for that site. Unsupported domains receive no AnyWebMCP site code.

  2. The site package registers its tools

    Each integration defines tool names, descriptions, input schemas, and website-specific behavior. It registers them with the page through document.modelContext.

  3. The agent discovers the registered interface

    A WebMCP-capable agent sees the registered tools and calls the one that matches the task. The tool returns the fields the agent needs for the task.

  4. The tool works in the current session

    Tools can use the rendered page, same-origin requests, public APIs, or visible controls. On signed-in sites, they use the browser session you already opened rather than asking for separate credentials.

Boundaries

Each site has an explicit integration

AnyWebMCP supports the websites listed in the catalog. Each site package defines its own tools, page requirements, and safety behavior. Some tools are read-only. Drafting tools prepare content for your review, and you submit it yourself.

Explore

See what is available

Browse each supported website and its available tools.

Open the website catalog →
Install

Choose your setup

Compare the available browser paths and select the one that fits your environment.

Go to Getting started →