Home / Projects / IlliniClaw
IlliniClaw

Most students never email the TA at 11 p.m. when they’re stuck—but they’ll message a WhatsApp bot. And most TAs spend office hours answering the same five questions every week. IlliniClaw is an open architecture spec for course-specific WhatsApp teaching assistants that absorb the FAQ load and route hard questions to humans. In production at Gies for BADM 554 with 36 active students.

36 Active Students
109+ Sessions & Counting
97+ Hours Engagement
View Spec on GitHub

How It Works

Feed the spec to an AI coding agent. It builds the entire system for your course.

1

Student Q&A via WhatsApp

Students join their course on WhatsApp and ask questions anytime. AI answers using RAG over course materials—lecture notes, assignments, textbooks—with conversation context maintained across 2-hour sessions.

  • Instant answers grounded in course materials
  • Context-aware conversations with session memory
  • Canvas LMS integration for real-time course data
2

Professor Dashboard

A web dashboard lets professors manage their courses, send broadcasts, and create drip campaigns. JWT authentication ensures only authorized faculty can access course controls.

  • Course management and student enrollment
  • Broadcast messaging to all enrolled students
  • 30 AI calls/hour rate limiting per professor
3

Drip Campaigns

Professors schedule multi-message drip campaigns to deliver content over time—test review materials, weekly study guides, or concept reinforcement.

  • Scheduled multi-message sequences
  • Customizable per course and topic
  • 3 campaigns deployed so far, more being added

Technical Architecture

The spec produces a production-ready system with these components

🔧

Backend

  • Node.js 20+ with ES Modules
  • Express web framework
  • PostgreSQL database
  • PM2 process management
🤖

AI System

  • Claude or OpenAI for student Q&A
  • RAG over course materials
  • 2-hour conversation session windows
  • Customizable persona per course
📱

WhatsApp

  • WhatsApp Business API integration
  • Canvas LMS integration
  • Drip campaign scheduler
  • Broadcast messaging
🔒

Security

  • JWT authentication (7-day expiry)
  • bcrypt password hashing (12 rounds)
  • VPN-ready dashboard access
  • Rate limiting (30 AI calls/hour)

BADM 554 Experiment

IlliniClaw is being actively tested with a live deployment for BADM 554 at Gies

Ongoing Deployment

The BADM 554 bot is actively serving 36 students in Enterprise Database Management, with 97+ hours of engagement across 109+ sessions so far and growing as the semester continues.

Drip Campaigns

Three drip campaigns deployed so far: Test 1 review (12 messages), Week 7 content (6 messages), and Week 8 content (10 messages). More campaigns are being added as the course progresses.

Open Spec, Private Code

The BADM 554 bot is a private implementation. The IlliniClaw spec is the open-source blueprint that lets anyone replicate the system for their own course.