Multi-Agent AI for Education

Student-led development at Gies College of Business exploring how intelligent agent teams can transform learning experiences

Why AgentLab?

🎓

Student-Led Innovation

Real students solving real educational challenges through hands-on development of AI systems

🤖

Multi-Agent Focus

Specialized AI teams working together, not just single chatbots. Each agent has domain expertise

📊

Evidence-Based

Grounded in educational theory and practical testing with measurable learning outcomes

Our Projects

Building production AI systems that bridge research and real classroom use

💡

VentureBots

Flagship

AI-powered entrepreneurship coaching that guides students from problem discovery to pitch-ready presentation using 5 specialized agents.

5 Agents 25 Students 90% Satisfaction
Google ADK Python React Streamlit
🧭

PathShaper

Adaptive Learning

Learning 3.0 architecture where faculty define course intent, multi-agent swarms generate personalized content, and faculty approve before deployment.

186 Concepts 39 Content Items 8 Weeks
Next.js 15 React 19 TypeScript Vercel

Canvas MCP

Infrastructure

MCP server providing 90+ tools and 8 agent skills for Canvas LMS. Enables 40+ AI coding agents to manage courses through natural language.

90+ Tools 8 Skills 290+ Tests
Python FastMCP PyPI Canvas API

UniQuick

Zero-Config Backend

A "deploy in one prompt" backend for AI-built web apps at Illinois. Ask your AI coding agent for an app and it goes live on a campus URL in seconds—Illinois sign-in, shared data, file uploads, realtime, and AI already wired in. No accounts, no API keys, no servers to manage.

1 Prompt to Live 5 Built-in Services Live on Azure
Fastify TypeScript Azure PostgreSQL MCP Entra SSO
💬

IlliniClaw

WhatsApp AI Blueprint

Open architecture spec for WhatsApp AI teaching assistants. Feed it to any AI coding agent and get a working bot for your course. Currently being tested with 36 students in BADM 554.

36 Students 109 Sessions 97+ Hours Engagement
Node.js Claude AI WhatsApp PostgreSQL Canvas LMS
🔎

GiesClaw

Autonomous Research

Autonomous business research platform where AI agents investigate companies using 13 specialized skills and publish structured analysis to a community-driven web platform. Humans and agents collaborate across finance, strategy, and more.

13 Agent Skills 6 Communities Live Platform
Python Next.js SEC EDGAR Yahoo Finance

Inquiring Agents

AI Ethics · Multi-Agent Debate

Five philosopher-agents debate AI ethics dilemmas through Churchman’s Inquiring Systems. Students pick a topic, watch streaming debates across 3 rounds, and inject their own perspectives. A coordinator synthesizes all positions into policy recommendations.

5 Philosopher Agents 3 Debate Rounds 8 Built-in Dilemmas
OpenAI GPT-4o Cloudflare Workers Server-Sent Events
🧠

The Cognitive Swarm

Real-Time Brainstorming

Real-time multimodal brainstorming app where groups contribute ideas through voice, see them cluster in 3D, and converge through quadratic voting—powered by Google Gemini.

3D Visualization Live Voice Anchor QV Quadratic Voting
React TypeScript Socket.IO Google Gemini
💫

MindForum

Shared AI Workspace

A shared AI workspace for group brainstorming. Create a room, share the link, upload documents, and chat together with an AI collaborator that only speaks when mentioned with @ai—keeping humans at the center of the conversation.

Group Workspace @ai Mention to Invoke PDF/DOCX Context
Next.js 15 TypeScript OpenAI API SSE
🎟️

HackClaw

AI Hackathon Assistant

Discord + Email assistant for the Gies AI for Impact Challenge hackathon. Routes questions through topic-specific brain files, manages a mentor help-ticket queue with claim/resolve buttons, and grows its solutions KB automatically as tickets resolve.

24h Coverage KB Self-Building 2 Channels
Python Discord.py OpenAI API Embeddings

Text-2-SQL Agent

Natural Language to SQL

Production-grade natural-language-to-SQL agent competing on AgentBeats. Decomposes questions, writes SQL, scores itself across 7 dimensions, and retries with targeted feedback when quality is low. Multi-step plans feed predecessor results into later queries.

7 Scoring Dimensions A2A AgentBeats Multi-Step Plans
Python LangGraph GPT-5 SQLite

ProgramOS

Spec for AI Program Coordinators

An AI program coordinator that answers stakeholder questions on email, Telegram, Teams, and web chat—and writes every decision into a shared, human-readable program folder. Start on a laptop in an afternoon, graduate to a server when you’re ready. Open recipe; you and your coding agent build it for your program.

1 Afternoon to First Bot 4+ Channels 0 Lines of Code to Copy
NanoClaw Claude Code Docker Spec-Driven
View All Projects

Latest from AgentLab

View All Updates →
June 14, 2026 · Release

UniQuick: Deploy an AI-Built App in One Prompt

A zero-config backend for AI-built web apps at Illinois—ask your agent for an app and it goes live on a campus URL in seconds, with sign-in, data, uploads, realtime, and AI already wired in...

Read More →
April 25, 2026 · Release

ProgramOS: An Open Spec for AI Program Coordinators

A spec, not a fork. Hand SPEC.md to your coding agent along with a fresh NanoClaw clone, and you get a multi-channel program coordinator covering email, Telegram, Teams, and web chat—extracted from a production deployment...

Read More →
April 21, 2026 · Release

MindForum Launches a Shared AI Workspace for Groups

A link-gated room where a group chats together with an AI collaborator that stays silent until mentioned with @ai—keeping humans at the center of the conversation...

Read More →