PantryPilot

A meal-planning and shopping-list app built for The WebMCP Challenge.

PantryPilot is an ordinary, useful web app — plan a week of dinners, respect dietary restrictions, track pantry staples, get an auto-generated shopping list — that becomes something new once an AI agent can drive it directly. Instead of an agent scraping the DOM or simulating clicks, this page exposes its real capabilities as WebMCP tools, so a human and an agent can plan the same week together, live, in the same shared state.

!!! quote “” “You have chicken, no dairy, and 20 minutes on Tuesdays. Plan my week and build the shopping list” is a sentence an agent can now act on, not just answer.

Try the live demo{ .md-button .md-button–primary } How to use it{ .md-button } See it on Devpost{ .md-button }

Why this fits the challenge theme

The theme asks for apps that are “meaningfully enhanced when both humans and AI agents can use them collaboratively.” PantryPilot is built so that:

Where to go next

Page What’s there
How to Use Walk through the live demo as a human, then as an agent — with screenshots and exact steps to enable WebMCP yourself
WebMCP Tools All 19 registered tools, their JSON Schemas, and what each one actually enforces
Architecture Diagrams: the human/agent/shared-core system design, the module dependency graph, and a full tool-call sequence
Development Run it locally, run the test suite, project structure, security notes

License

MIT — see LICENSE.