Campus Buddy

A voice-first university schedule assistant, built as a simulated Alexa+ experience for the Build, Ship, Shape: Amazon Developer Hackathon (Alexa+ track), in English and Bangla (বাংলা). It also ships as a self-hosted MCP server (Streamable HTTP, spec 2025-11-25) so any MCP client or agent can use the same assistant.

Live demo: https://iamsajid0010.github.io/campus-buddy/ (use Chrome or Edge for voice)

Students can ask, by voice or text:

Campus events: it also answers questions about seminars, workshops, contests, hackathons and fests: when, where, who organizes them. It checks each event against the student's class routine and warns about clashes, e.g. "clashes with your Electronic Devices and Circuits Lab".

Project description

Campus Buddy is a conversational schedule assistant for university students, built as a simulated Alexa+ experience in a web app. Instead of opening a portal or scrolling through screenshots of a routine, a student simply asks, "When is my next class?" or "Which assignments are due this week?", and the assistant answers out loud with the time, room, and teacher. It also tracks campus events such as tech seminars, workshops, contests and hackathons, tells students when and where they are, and warns them when an event clashes with one of their classes.

It works in English and Bangla: ask in Bangla (by voice or text) and it answers in Bangla, with Bangla times ("সকাল ৮:৩০"), Bangla digits and day names, and spoken replies when the browser has a Bangla voice. All schedule data lives in a single editable data file, so any student can make it their own in minutes, with no database or account. The same logic is also exposed as an MCP server with eight read-only tools, so an agent can answer "do I have a clash with the Tech Talk?" through the open standard that Alexa+ integrations use. Campus Buddy shows how a voice assistant can remove everyday friction for students who juggle many classes and deadlines, in their own language.

Related MCP server: MCP-Server-CollageAI

How it works

• index.html - the page: orb, chat, microphone button.

• app.js - the interface. Uses the browser's Web Speech API for voice input (speech recognition) and spoken replies (speech synthesis). Replies come back in the language of the question.

• engine.js - the assistant's logic. Understands a question with keyword and intent matching, then looks up the answer in the data and builds a reply in English or Bangla. Pure logic, so it runs in the browser and in Node.

• data.js - your schedule (classes, assignments, exams, campus events). The class routine comes from a real department routine; assignments, exams and events are sample data to replace. Fields starting with bn hold the Bangla names.

• mcp/ - the MCP server (see below). It imports engine.js and data.js, so the web app and the MCP tools always agree.

The web app needs no server, no database and no API keys. Everything runs in the browser.

Run the web app

1. Download or clone this repository.

2. Open index.html in Google Chrome or Microsoft Edge (voice input needs one of them).

3. Click the microphone and speak, or type a question. Pick বাংলা in the language menu for Bangla.

Optional: serve it locally with python -m http.server and open http://localhost:8000.

MCP server (self-hosted, Streamable HTTP)

cd mcp
npm install
TZ=Asia/Dhaka npm start        # http://127.0.0.1:3333/mcp   (health check: /health)

• Transport: Streamable HTTP, endpoint POST /mcp, stateless. Built with the official @modelcontextprotocol/sdk; the protocol version 2025-11-25 is negotiated (older clients are supported by the SDK).

• Binds to 127.0.0.1 by default and rejects browser requests from foreign origins (DNS-rebinding guard). Set HOST, PORT or ALLOWED_ORIGINS to change that.

• All tools are read-only and return both readable text and structured content (say, items, language).

Every tool takes an optional language (en or bn) and an optional now (ISO date-time, for testing).

Try it with plain HTTP:

curl -s localhost:3333/mcp -H 'content-type: application/json' -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"next_class","arguments":{"language":"bn"}}}'

Or point any MCP client at http://127.0.0.1:3333/mcp (for example the MCP Inspector: npx @modelcontextprotocol/inspector, transport "Streamable HTTP").

Tests

node test/engine.test.js        # engine: every suggestion in English and Bangla, clash detection
cd mcp && npm install && npm test   # MCP: starts the server and calls every tool with the official MCP client

Use your own schedule

Open data.js and replace the sample courses, classes, assignments and exams. Days use 0 = Sunday ... 6 = Saturday, and times use 24-hour "HH:MM". Add the bn... fields if you want Bangla replies to use Bangla names; without them the English text is used.

Limitations (honest notes)

• This is a simulation of an Alexa+ experience; it does not run on an Alexa device. The MCP server follows the open MCP spec, but it has not been registered with or tested on a real Alexa+ device.

• Question understanding uses keyword matching, so unusual phrasings may not be understood.

• Bangla spoken replies need a Bangla voice installed in the browser (Microsoft Edge often includes one). Without it, Bangla replies are shown as text and the app says so once.

• Bangla course and teacher names in data.js are transliterations; check the spellings.

• Assignments, exams and events are sample data. The MCP server computes sample dates when it starts.

License

MIT, see LICENSE.