13 days ago
Meet the winners 🏆
The moment you're been waiting for: judging is complete, and the winners of the WebMCP Challenge are here!
First, thank you to everyone who built something! Over 7,000 people registered, and nearly 2,500 projects came in from more than 100 countries in just 10 days. Most of you were new to WebMCP when you started, and you still shipped real, working apps. You learned an emerging standard fast and built something real with it. The project gallery is full of great work, so take a look at what everyone built.
The Winners 
MASIL — Reconnecting Korean Elders to Creative Life
A creative space for Korean elders who live alone and can't easily leave home. Their agent adds calligraphy references and Janggi moves through WebMCP, and the elder finishes the work themselves.
Alza
A floor-plan editor that shares its tools with agents. Give the agent a photo of a plan and one real measurement, and it traces the walls, rooms, doors, and furniture into a 3D model you can walk through.
ArchMorph
A home design studio where you and an agent work on the same building in a floor plan, a 3D view, and a walk-through mode. The agent uses 57 typed tools to edit the design and catch layout problems early.
Aisle - AI Wedding Seating
A wedding seating planner where you and an agent share one live seating chart. The agent arranges tables, fixes conflicts, and explains its choices, and it keeps any guest placements you pin.
Roque Nights
A stargazing planner built by an engineer at the Gran Telescopio Canarias. The agent ranks the darkest nights, picks what to observe, and points the sky map while you watch.
Observatory
A fantasy map builder for writers and game designers. Ask for "a lone tower in a dense forest," and the agent builds editable terrain, buildings, and labels, not a flat image.
Bouquet Studio
A flower shop where you describe who a bouquet is for and how it should feel. The agent builds a first arrangement from real products, and you shape it together.
JupyterLite WebMCP
An agent that works right inside a browser-only Jupyter notebook. It reads your unsaved cells and exact selections, runs code in the kernel you share with it, and marks every change in place.
Faraday
A browser viewer for CT and MRI scans. The agent finds and measures regions without ever receiving the scan, and exporting findings needs a person's approval.
Mandate
A permission layer that lets you give an agent access to specific records and fields for a set time. The agent can stage changes but can't commit them, and its access ends when time runs out or you revoke it.
WebMCP is still early, and you're now some of the people who know it best. Keep building, and share what you make in the OpenAI Discord.