Skip to main content
OpenAI

The WebMCP Challenge

Meet the winning teams exploring what’s possible when people and agents use the web together.

Make the web work with agents

WebMCP ↗⁠(opens in a new window) is an experimental open standard that lets websites expose structured tools agents can use directly. Instead of leaving agents to guess their way through your UI, you define exactly how they can use your app, so they complete tasks faster, more accurately, and more reliably.

The WebMCP Challenge invited builders to create apps that become meaningfully better when people and their agents can use them together.

Supported by

Winning teams

MASIL lettering beside calligraphy tools.

MASIL⁠(opens in a new window)

A creative space that helps Korean elders practice calligraphy and play Janggi with an agent through voice and direct interaction.

Alza interface showing a furnished three-dimensional floor plan.

Alza⁠(opens in a new window)

A floor-plan editor where people and agents draw, check, and furnish a shared 2D plan, then explore it in 3D.

ArchMorph architectural floor plan and three-dimensional building model.

ArchMorph⁠(opens in a new window)

A browser-based architecture studio where people and agents design and inspect a home together through floor plans, 3D views, and walk-throughs.

Aisle project title with a gold floral emblem.

Aisle⁠(opens in a new window)

A shared wedding seating canvas where people and agents arrange guests, check seating constraints, and resolve conflicts.

Roque Nights title over a star-filled night sky.

Roque Nights⁠(opens in a new window)

A night-sky planner where people and agents choose observing windows and targets, then review a shared plan on an interactive timeline.

Mandate permission interface with temporary authorization details.

Mandate⁠(opens in a new window)

A delegation layer that lets people give agents temporary access to selected records and fields, review staged changes, and revoke access.

Observatory fantasy map canvas showing a ruined stone tower in a forest clearing.

Observatory⁠(opens in a new window)

A shared fantasy map editor where people and agents build terrain, rooms, and objects on an editable canvas.

Bouquet Studio title beside a wrapped bouquet of pale flowers.

Bouquet Studio⁠(opens in a new window)

A shared bouquet editor that turns a person’s occasion, mood, and budget into flower arrangements they can see and refine with an agent.

Faraday research viewer displaying brain scan slices.

Faraday⁠(opens in a new window)

A browser-based CT and MRI viewer that lets agents navigate volumes and work with measurements while scans stay local. For research and education only, not diagnosis.

JupyterLite notebook interface with the words The agent edits. You see every change.

JupyterLite WebMCP⁠(opens in a new window)

A JupyterLite extension that lets agents work with live notebook cells, selections, and the browser’s shared Python kernel, with inline review of proposed edits.

What will you build with WebMCP?

Explore a few agent-native apps from our team for inspiration, or browse the WebMCP showcase ↗⁠(opens in a new window) for more ideas.

Meet the judges

Headshot of Sarah Drasner, Distinguished Engineer, Chrome, Google.

Sarah Drasner

Distinguished Engineer, Chrome, Google

Headshot of Andrew Galloni, VP Research & Innovation, Cloudflare.

Andrew Galloni

VP Research & Innovation, Cloudflare

Headshot of Jude Gao, Member of Technical Staff, Vercel · Next.js Core Team.

Jude Gao

Member of Technical Staff, Vercel · Next.js Core Team

Headshot of Ilya Grigorik, Distinguished Engineer, Shopify.

Ilya Grigorik

Distinguished Engineer, Shopify

Headshot of Alex Nahas, Creator of MCP-B.

Alex Nahas

Creator of MCP-B

Headshot of Sean Roberts, VP of Applied AI, Netlify.

Sean Roberts

VP of Applied AI, Netlify

Headshot of Justin Rushing, Browser Agent Lead, OpenAI.

Justin Rushing

Browser Agent Lead, OpenAI

Build with WebMCP

Explore what’s possible when people and agents can use the web together.