Personalized Learning with AI

Open Knowledge Tracing Service

A tool that supports accurate diagnosis and effective learning by providing Knowledge Tracing—the core component of Intelligent Tutoring Systems—as a Model Context Protocol server.

Carnegie Mellon University HCII
Pittsburgh, PA

The Challenge

No Learning Science Integration

Current LLM tutors lack proper integration with learning science principles for personalization.

Domain-Specific Limitations

Systems rely on domain-specific interfaces, limiting scalability and accessibility across subjects.

Lack of Research Data

Insufficient datasets to support research on improving LLM-based educational experiences.

Our Solution

We develop an MCP (Model Context Protocol) server that enables existing LLM services like ChatGPT to be adapted into personalized tutoring systems. Our Knowledge Tracing methodology:

  • Can be applied to newly generated questions without relying on historical data
  • Works regardless of memory type (declarative or procedural)
  • Operates effectively in natural language interaction environments

Key Features

Progress Metrics

Intuitive representation of learner progress with target-time optimization for exams and interviews.

ChatGPT Apps Integration

Seamless integration with ChatGPT Apps and Claude Artifacts for familiar user experiences.

Open Dataset

Contributing valuable research data for the learning science community.

Our Goal

30K
Total Users
5K
Monthly Active Users
200K+
Learning Interactions
300M
Potential Reach via ChatGPT