Skip to main content
Glama
tejaswi-raghav

PantryPilot MCP Server

PantryPilot — Alexa+ Experience

PantryPilot is a polished, browser-based simulation of a brand experience for Alexa+. A customer asks for a weeknight dinner, Alexa+ coordinates brand tools, and the customer can tune servings or approve missing ingredients without leaving the conversation.

Live demo: pantrypilot-alexa-plus.vercel.app

This submission follows the “new to MCP” path in the project brief: it is a self-contained simulated experience, not a production Alexa integration. The interaction model is informed by the MCP Apps Agent Skills guide: structured tools drive the same state as the visible interface, tool activity is transparent, and consequential actions require explicit confirmation.

What the demo shows

  • A realistic Alexa+ request-to-result conversation

  • A visual dinner recommendation based on pantry, diet, and time constraints

  • Adjustable servings with live ingredient scaling

  • Three selectable pantry-first meal options

  • An approval-gated shopping list with removable items and feedback

  • A guided cooking mode with step progress and working timers

  • Conversational and one-tap plan refinements

  • A live brand-tool trace for read_pantry, find_recipe, and build_cart

  • An inspectable JSON result with a clear purchaseMade: false boundary

  • Responsive desktop and mobile layouts

  • Browser-native WebMCP tools that agents can discover and call

  • A production-hosted Streamable HTTP MCP endpoint at /mcp

Related MCP server: Gamito

Run locally

Requirements: Node.js 22.13 or newer and npm.

npm install
npm run dev

Open http://localhost:5173.

Create a production build with:

npm run build

The repository is a standard Next.js project and can be imported directly into Vercel with no environment variables.

Streamable HTTP MCP server

The same Vercel deployment hosts a stateless MCP server at:

https://pantrypilot-alexa-plus.vercel.app/mcp

It uses the official MCP TypeScript SDK, negotiates protocol version 2025-11-25, supports Streamable HTTP POST requests and optional GET streams, validates browser Origin headers, and scales safely across serverless instances. No API key is required because the server exposes only simulated demo data.

Server-side capabilities:

Capability

Purpose

read_pantry

Read the simulated household pantry

find_recipes

Rank pantry-first recipes by time and preferences

get_recipe

Read a complete guided-cooking plan

build_shopping_list

Stage missing ingredients after confirmed: true; never purchases

pantry://current

Read the pantry as an MCP resource

Test protocol negotiation:

curl https://pantrypilot-alexa-plus.vercel.app/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  --data '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"curl","version":"1.0.0"}}}'

Agent-facing tools

Browsers that support the proposed imperative WebMCP interface can discover five page-scoped tools:

Tool

Purpose

Side effect

read_current_recipe

Read the visible recipe, time, servings, and missing-item count

None

select_recipe

Select one of the three visible recipe options

Updates the active meal

configure_recipe_servings

Set servings from 1–12

Updates the visible recipe card and ingredient quantities

add_missing_items_to_list

Add missing ingredients after confirmed: true

Updates the simulated list; never purchases

start_guided_cooking

Open the selected meal at step one

Opens the visible guided-cooking dialog

Each tool validates its input and updates the same React state as the on-screen controls. Unsupported browsers simply ignore the optional registration and the visual demo continues to work.

Interaction checklist

  1. Switch among the three meal options or use a quick refinement.

  2. Use the plus/minus controls to scale servings and ingredients.

  3. Add missing items, then remove individual items from the list if needed.

  4. Open Start guided cooking and move through the recipe steps or run a timer.

  5. Open Inspect result to see the structured response data.

Project structure

app/
  page.tsx       Experience UI, state, and WebMCP registrations
  globals.css    Responsive visual system and interaction states
  layout.tsx     Page metadata and document shell
public/
  grain-bowl.jpg Optimized original recipe photography
  favicon.svg    PantryPilot mark

The project uses Next.js 16, React 19, TypeScript, Tailwind CSS 4, and Lucide icons. It requires no API key, database, account, or external service.

Accessibility and trust

  • Semantic landmarks and useful accessible names

  • Visible keyboard focus states

  • Reduced-motion support

  • Responsive behavior down to mobile widths

  • No purchase, checkout, or external transmission in the simulation

  • Explicit confirmation before updating the shopping list

From simulation to production

For a production Alexa+ submission, replace the in-page simulated tools with a self-hosted Streamable HTTP MCP server using specification 2025-11-25 or later, preserve the same tool boundaries, add authentication and durable household state, and test with the current Alexa+ developer preview requirements.

Asset note

The food photograph was generated specifically for this project. No external runtime assets are fetched.

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    C
    maintenance
    Household-aware kitchen brain for AI agents: manage pantry inventory with freshness tracking, shopping lists, recipe collections with cook notes and per-diner ratings, dietary profiles with allergen safety, and kitchen equipment — all through 27 tools with OAuth 2.1 authentication. Includes a free tool for ingredient-based recipe generation without an account (accounts are free!).
    MIT
  • F
    license
    B
    quality
    D
    maintenance
    Enables AI agents to generate budget-disciplined, allergy-safe weekly meal plans, shopping lists, and meal swaps using a fully local deterministic engine.
    20
    -
  • A
    license
    Not graded
    quality
    B
    maintenance
    Enables AI assistants to read current pantry inventory and low-stock shopping needs, export a shopping queue, and perform idempotent inventory adjustments and catalog link management through remote MCP tools.
    MIT
  • F
    license
    Not graded
    quality
    B
    maintenance
    Enables MCP-compatible clients like ChatGPT and Claude to manage a household's shared food inventory, meal plans, recipes, and preferences through natural language, with atomic updates and audit logging.
    -