Syllabus.AI

Inspiration

As engineering students, we were drowning in scattered information — deadlines buried in WhatsApp groups, notes spread across Drive folders and physical notebooks, and no single place to just ask what's due or what a topic means. We kept wishing we could just talk to an AI that actually knew our syllabus, not a generic chatbot giving textbook answers. When we learned about MCP — a standard that lets AI assistants plug into real data sources — we realized we could finally build that.

What it does

Syllabus.AI is an MCP server that connects an AI assistant directly to a student's real study data — deadlines, notes, and topics — stored in a simple Google Sheet. Instead of manually searching for "what's due this week" or "summarize my OS notes," students just ask their AI assistant in plain language, and it pulls live, accurate answers from their actual coursework data.

How we built it

We built the MCP server in Python using FastMCP, exposing two core tools: get_deadlines and summarize_notes. These tools read directly from a connected Google Sheet via the Sheets API. We connected the server to Claude Desktop as our MCP client for the demo, so any query typed into the chat triggers a real tool call, fetches live data, and returns a grounded answer instead of a hallucinated one.

Challenges we ran into

Getting Google Sheets API authentication working smoothly under a tight timeline was our biggest hurdle. We also had to think carefully about tool design — making sure each MCP tool had a clear, single responsibility so the AI could reliably choose the right one instead of getting confused between overlapping tools. Debugging the MCP client-server handshake with limited documentation examples also took longer than expected.

Accomplishments that we're proud of

We went from zero MCP experience to a working, demoable integration in a very short window. Seeing the AI answer a real question using our actual deadline data — instead of a generic response — was the moment it clicked that we'd built something genuinely useful, not just a tech demo.

What we learned

We learned how MCP standardizes AI-to-tool communication, why that matters compared to building one-off API integrations, and how to design tools an LLM can reliably choose between. We also learned the value of scoping down aggressively — cutting our tool list from five ideas to two so we'd actually have something working to show.

What's next for Syllabus.AI

Next, we want to add more tools — explain_topic for on-demand topic explanations, and integration with sources like class timetables and Unstop for deadline tracking beyond just coursework. We'd also like to move from a Google Sheet backend to a proper database and support multiple students with their own private data scopes.