Explainer mini-site
An interactive comparison between DOM-driven agent interaction and pages that publish explicit, structured WebMCP tools.
Browse 50 curated demos, libraries, tools, and official guides for trying and building with WebMCP. What is WebMCP?
50 resources
An interactive comparison between DOM-driven agent interaction and pages that publish explicit, structured WebMCP tools.
A restaurant booking flow whose annotated form lets an agent validate details and submit a reservation through a declared tool.
A React travel search interface that exposes flight criteria and results as callable tools while keeping the UI in sync.
A hands-on pizza configurator where imperative tools let an agent choose the size, style, sauce, toppings, and extras.
A small multi-page adventure that demonstrates how declarative form tools and imperative JavaScript tools can work together.
A maze game designed to be started, inspected, and solved entirely through an agent calling page-level tools.
A cinema checkout prototype where agents can change location, filter films, choose a showing, and begin a ticket order.
A simulated retail account that declares tools for looking up recent deliveries and starting a return from the same page.
A polished hotel finder that combines annotated forms with registered functions for searching properties and comparing amenities.
An Angular sporting-goods shop that lets agents search its catalog and operate the cart through both WebMCP API styles.
A React storefront that makes product discovery, cart management, and the purchase path available to an agent.
A compact flight finder published in parallel imperative and declarative versions for easy implementation comparison.
A minimal CodeSandbox example in which a page tool switches the displayed image between a cat and a dog.
A chessboard that publishes tools for reading the position, listing legal moves, moving pieces, promoting pawns, and restarting.
A tutorial and sample booking app that contrasts agent-native tool calls with screen scraping, including a token-use case study.
A live service whose page tool launches an agent-readiness review covering discoverability, structured data, content, and SEO.
A blackjack table where separate player, opponent, and dealer agents repeatedly observe state and act through described tools.
A browser-side bridge that projects tools from a remote MCP server into the page API so in-browser agents can call them.
A diagramming pipeline that turns a natural-language request into Mermaid, validates it, and renders the result in Excalidraw.
An agent-operated architecture canvas with tools for adding nodes, connecting edges, labeling relationships, and applying layout.
A premium coffee shop prototype with agent-friendly catalog navigation, repeat ordering, and product specification lookup.
A property map whose registered tools let an agent change location, price, amenity, and map filters from a request.
An Angular leather-goods store with declared tools for product search, policy checks, and shopping-cart actions.
A configurable home dashboard where tools reveal and rearrange controls for cameras, lighting, climate, and energy.
A searchable directory of WebMCP-enabled sites with a JSON discovery API and its own tools for querying and submitting entries.
A Gemini-powered controller that accepts natural language and invokes tools offered by another WebMCP-enabled page.
A statistics workbench that can fetch JSON-stat datasets, filter dimensions, build cross-tabs, and export the selection to CSV.
A no-sign-in publishing service whose page tool accepts generated HTML and returns a private share link plus an ownership claim link.
A casual guessing game that uses the browser Prompt API to generate words and exposes the complete game loop through WebMCP.
A browser-based 3D studio where an agent can inspect scenes, edit geometry and materials, and render assets while the user watches in the same viewport.
A local note editor where an agent can create notes, read comments, and respond under its own identity inside the document.
A shared crossword workspace where an agent can propose themed words, build a grid, refine clues, and solve alongside the user.
A 12-voice beat machine whose tools let an agent compose patterns, tune groove and sounds, and iterate while playback stays editable.
An editable trip planner where an agent turns context into an hourly itinerary and map, then responds to comments and revisions.
A browser photo editor where an agent can adjust exposure, color, composition, and image state while the user reviews each change.
A shared weekly meal planner that lets an agent coordinate recipes, groceries, and preferences without overwriting the user's edits.
A 3D puzzle cube whose tools expose every face and accept move sequences, allowing an agent to plan and animate a solution.
A greeting-card canvas where an agent can supply artwork and copy, then refine format, messages, and envelope details with the user.
A grocery storefront where an agent can search 110 products, inspect items, and manage a shared cart through structured page tools.
A one-script integration that detects common search, cart, and form handlers, then registers matching WebMCP tools without a framework.
A React hook that registers a page tool when a component mounts and automatically handles updates and unregistration.
A TypeScript declaration package for applications that use the experimental browser WebMCP interfaces.
A Chrome extension for discovering a page's WebMCP tools, viewing their input schemas, invoking them manually, and diagnosing failures.
An agent skill that inventories an existing app, proposes a tool surface, implements approved tools, and verifies them in a browser.
A tutorial template that gives an 11ty comic archive a website, an MCP server, and browser tools for reading and navigation.
The canonical Web Machine Learning Community Group draft defining declarative form tools and the imperative model-context API.
Chrome's practical introduction to WebMCP, including setup, API choices, security constraints, examples, and testing guidance.
The GoogleChromeLabs-maintained index that collects live WebMCP demonstrations, reusable libraries, and developer utilities.
A command-line evaluation harness for checking whether an interactive model selects the expected tool and produces valid arguments.
A lightweight compatibility layer used by the GoogleChromeLabs demos to emulate WebMCP tools and related CSS states in other browsers.