GoogleChromeLabs/webmcp-tools

What it solves

This project provides a suite of developer tools and demonstrations to help developers adopt the WebMCP API, which allows web pages to expose structured tools to AI agents, enabling them to interact with web applications programmatically rather than relying on site scraping.

How it works

The repository contains three main components:

1. Developer Utilities: Includes a Chrome Extension for inspecting and debugging WebMCP tools, a CLI tool for evaluating LLM tool-calling capabilities, and a JavaScript polyfill for browser compatibility.
2. Declarative Tools: Demos showing how web pages can define tools using a structured, declarative approach.
3. Imperative Tools: Demos showing how agents can control UI components (like flight searches, hotel bookings, and smart home dashboards) through imperative tool registration.

Who it’s for

Web developers and AI agent developers who want to integrate their websites with the Model Context Protocol (MCP) for the web, allowing AI assistants to navigate and operate their sites more reliably.

Highlights

• Model Context Tool Inspector: A browser extension for visualizing input schemas and debugging connection issues.
• WebMCP Evals: A CLI tool to verify if agents correctly call tools based on user inputs.
• Diverse Demo Suite: A wide range of implementations across various industries (travel, e-commerce, smart home, and hospitality) using both React and Angular.
• WebMCP Page Agent: A Gemini-powered demo that allows natural language control of any WebMCP-enabled website.