DayFlow — Alexa+ Hackathon Project

DayFlow is a personal productivity assistant prototype built for the Alexa+ track of the Build, Ship, Shape: Amazon Developer Hackathon. It combines a polished browser-based simulated assistant with a self-hosted Model Context Protocol (MCP) server over Streamable HTTP.

Important: this is an independent hackathon prototype, not an official Amazon product and not connected to an Alexa+ account. The browser assistant uses deterministic intent routing (not an LLM) so its actions are reliable and demonstrable. The MCP endpoint is real and exposes callable tools for compatible MCP clients.

Features

• Browser demo with task overview, task creation, completion, and a focus planner.

• Five real MCP tools: dayflow_add_task, dayflow_list_tasks, dayflow_complete_task, dayflow_plan_focus, and dayflow_daily_briefing.

• MCP Streamable HTTP endpoint at POST /mcp, implemented with @modelcontextprotocol/sdk.

• Shared core functions: the browser API and MCP tools operate on the same task state.

• No API key, account, or paid service needed to run the local demo.

• Sample data lives in memory and resets when the server restarts. This keeps setup quick for the hackathon; persistent storage is a future enhancement.

Related MCP server: Chisel Planner

Requirements

• Node.js 20 or newer

• npm

• VS Code (recommended)

Run locally

1. Extract this project and open the DayFlow-AlexaPlus folder in VS Code.

2. Open Terminal → New Terminal.

3. Install packages:

   npm install

4. Start the server:

   npm start

5. Open http://localhost:3000.

6. MCP endpoint: http://localhost:3000/mcp.

7. Health check: http://localhost:3000/api/health.

For development with automatic restart, run npm run dev.

MCP integration notes

The server imports and calls McpServer and StreamableHTTPServerTransport from the official MCP TypeScript SDK. The endpoint is stateless and uses Streamable HTTP. Tool schemas are validated with Zod.

The five tools:

Smoke-test the endpoint

After starting the server, send an MCP initialization request using a REST client such as Thunder Client in VS Code or curl:

curl -i http://localhost:3000/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "MCP-Protocol-Version: 2025-11-25" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"dayflow-smoke-test","version":"1.0.0"}}}'

The server responds with the negotiated protocol version and capabilities. For a real MCP client, use its supported Streamable HTTP transport and point it to http://localhost:3000/mcp. Because this demo is stateless, each POST creates a fresh server/transport pair; client integrations should follow the SDK's stateless Streamable HTTP behavior.

Demo walkthrough (60–90 seconds)

1. Open the dashboard and point out the open-task summary.

2. Ask: “Plan my next 90 minutes.” Show the prioritized plan.

3. Ask: “Add task to prepare my hackathon demo.” Show the new task appear.

4. Mark a task complete and show the summary update.

5. Point to the five MCP tools and the /mcp endpoint.

6. Explain that a compatible Alexa+ experience can invoke the same tools through MCP; this web UI is a local simulator, not a real Alexa+ integration.

Submission checklist

• Verify current official rules on the Devpost page.

• Put this source in a GitHub repository.

• Add an open-source license (MIT is included as LICENSE).

• Test the app from a fresh install.

• Record a public YouTube/Vimeo demo under three minutes, in English.

• Add the repository and demo video links to Devpost.

• Complete product feedback for the MCP SDK and any other tools actually used.

• Only claim integrations you have actually tested. Do not claim the browser simulator is a live Alexa+ client.

Product feedback notes to customize honestly

• Tool used: @modelcontextprotocol/sdk with Streamable HTTP, plus Node.js/Express for the web app and REST demo API.

• What worked well: A shared set of functions makes it possible to demonstrate the same task actions in both the browser and MCP tools.

• What needs work: Stateless transport and client setup should be tested with the exact MCP host/client you intend to use; browser intent routing is intentionally limited and is not a general AI agent.

• Onboarding: Record the actual steps, errors, documentation pages, and time you experienced while setting up the SDK.

• Build with it again: Decide based on your real test experience and explain why.

Project structure

DayFlow-AlexaPlus/
├── public/
│   ├── app.js
│   ├── index.html
│   └── styles.css
├── src/
│   └── server.js
├── LICENSE
├── package.json
└── README.md

License

MIT. See LICENSE.