Protocol Engine: Built with @modelcontextprotocol/sdk (TypeScript) implementing standard JSON-RPC 2.0 contracts with dual Streamable HTTP and Server-Sent Events (SSE) transports.
5 Expressive MCP Tools: Designed strict JSON Schema contracts for get_tasks, add_task, create_study_plan, update_progress, and get_course_progress with rich parameter validation.
Adaptive Time Allocation Engine: Study sessions dynamically partition available time ( T_{\text{total}} ) into ( n ) deep-work sprints and ( (n-1) ) restorative intervals: $$T_{\text{total}} = \sum_{i=1}^{n} t_{\text{sprint}, i} + \sum_{j=1}^{n-1} t_{\text{break}, j}$$ Assignments are prioritized based on course weight ( W_{c} ), task urgency ( U_{t} ), and remaining time until deadline ( \Delta t ): $$\text{Priority Score} = \frac{W_{c} \times U_{t}}{\max(1, \Delta t_{\text{hours}})}$$
Local-First Database: Powered by Node 22 native node:sqlite in Write-Ahead Logging (WAL) mode with sub-millisecond execution times.
Web Companion Dashboard: Built with React 18, Vite, and Tailwind CSS, featuring an intelligent in-memory fallback layer for serverless cloud deployments on Vercel.

Challenges we ran into

Conversational Time Estimation & Granularity: Converting a student’s natural phrasing ("I only have an hour and a half before dinner") into structured, pedagogically sound focus cycles required rigorous MCP tool schema definitions and prompt alignment so Alexa's tool-selection model reliably returned valid numerical values.
Streaming Protocol & SSE Compliance: Implementing real-time synchronization between Alexa's remote invocation and our local web dashboard required building a dual-mode transport layer supporting both stateless POST requests (/mcp) and stateful Server-Sent Events (/mcp/sse).
Zero-Config Cloud Tunneling: Ensuring Alexa could communicate securely over HTTPS with our local development machine without requiring commercial domain registrations or credit cards was solved using Cloudflare Quick Tunnels (cloudflared).

Accomplishments that we're proud of

True Zero-Friction Workflow: A student can walk into their dorm room, never touch a keyboard or smartphone screen, and completely organize their evening's coursework and study timers purely by talking.
Instantaneous Response Latency: Thanks to native SQLite in WAL mode and efficient JSON-RPC dispatching, Alexa receives tool execution payloads in under 10ms.
Bridging Voice with Visual UI: Solving the "blind voice" problem by pairing Alexa's spoken dialogue with a synchronized, live companion dashboard that automatically visualizes active timers and course progress rings.

The true power of the Model Context Protocol (MCP): decoupling conversational AI reasoning from business logic creates an extensible architecture where voice assistants act as genuine autonomous agents rather than glorified smart switches.
Designing voice UX requires strict conciseness: spoken responses must convey priority and actionable numbers without overwhelming the student's cognitive load.
Local-first architectures offer unprecedented speed and data privacy for personal academic data.

📅 LMS Calendar Integration: Direct two-way synchronization with Canvas, Blackboard, and Google Calendar via additional MCP tool integrations.
🧠 Cognitive Fatigue Detection: Tracking task completion velocity over time to proactively suggest longer restorative breaks when focus efficiency drops: $$\text{Focus Efficiency } (\eta) = \left( \frac{T_{\text{verified focus}}}{T_{\text{scheduled}}} \right) \times 100\%$$
🔔 Proactive Audio Check-ins: Allowing Alexa to gently check in when a Pomodoro sprint completes to log immediate feedback on assignment difficulty.

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.