Inspiration
Between rushed morning commutes, sudden lab room shifts, and surprise assignment due dates, staying on top of college life is stressful. Regular voice assistants usually don't help much here because you have to say rigid, exact commands, and they don't know anything about your college's schedule. When I explored Alexa+ and the Model Context Protocol (MCP), I realized we could build an assistant that actually understands natural student questions and checks private college records on the fly.
What it does
Campus Academic Co-Pilot is a lightweight Python MCP server that provides three useful tools to Alexa+:
- get_daily_schedule(): Tells you your lectures, labs, and exact room numbers for the day.
- check_upcoming_deadlines(): Pulls upcoming project submissions and assignment dates so you don't miss anything.
- get_revision_quiz(topic): Gives quick active-recall quiz questions on topics like Python and AI to help revise before exams.
Instead of writing rigid intents, Alexa+ uses agentic reasoning to figure out what you need, calls the right tool in our Python backend, and gives a clean, conversational answer.
How we built it
- Wrote the backend in Python using the Model Context Protocol (
mcp/ FastMCP). - Configured tools using standard Python type hints and docstrings, which the LLM reads automatically to know when and how to call them.
- Ran the communication over standard I/O (stdio) via JSON-RPC.
- Managed and version-controlled the project through Git and GitHub.
Challenges we ran into
Since this was my first time working with MCP, shifting my mental model from traditional intent-slot skill design to autonomous agent tool-calling took some time to wrap my head around. Getting the Git credentials and remote push configured properly in VS Code also took a bit of troubleshooting, but working through it step by step got everything running cleanly.
Accomplishments that we're proud of
- Built a working MCP server from scratch and got all three tools responding properly.
- Implemented both zero-argument functions and parameterized functions (like passing the quiz topic dynamically).
- Pushed clean, working code to GitHub and recorded a live walkthrough demo within a tight deadline!
What we learned
- How MCP acts as a clean bridge between an LLM brain and local code/databases.
- How typing and docstrings in Python can directly power tool selection in Agentic AI.
What's next for Campus Academic Co-Pilot
- Connecting it with real university APIs or Google Calendar rather than mock data.
- Deploying the server onto AWS Lambda or Amazon Bedrock AgentCore for true 24/7 cloud availability.
Log in or sign up for Devpost to join the conversation.