Skip to content

Multi-agent AI tutoring architecture with an orchestrating Tutor Agent and specialist agents

Design · All designs
Evidence · 2 claims (2 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 2 claims rest on one study

Description

The article proposes a multi-agent architecture implemented on LangGraph in which a central Tutor Agent (GPT-4o) interprets student input and orchestrates other components via a ReAct-style framework. A Memory Dispatcher oversees conversations and controls memory modules, while a pipeline of agents (Research, Planning, Step Handling, Coding) builds courses, and auxiliary tools include a Symbolic Solver, Function Plotter, and Course Graph Drawer. The design prioritizes "deep understanding and independent problem-solving over direct answers" through guided, tool-assisted learning.

Design Implications

Context

Requirements

  • Underlying LLM capabilities: the article states system performance remains fundamentally dependent on underlying LLM capabilities and possible biases

Constraints

  • Personalization currently operates on a basic set of attributes
  • Pedagogical effectiveness of generated courses is yet to be evaluated empirically

Target Learners

  • students learning mathematics, including exam-preparation students

Learning Goals

  • deep conceptual understanding of mathematics topics
  • independent problem-solving
  • exam revision and targeted practice

Claims

Examples

-

Key Sources

  • Jarosław A. Chudziak and Adam Kostka. (2025). AI-Powered Math Tutoring: Platform for Personalized and Adaptive Education. arXiv preprint. https://arxiv.org/abs/2507.12484
Page metadata
Field Value
type design
id multi-agent-ai-math-tutor-architecture
title Multi-agent AI tutoring architecture with an orchestrating Tutor Agent and specialist agents
description The article proposes a multi-agent architecture implemented on LangGraph in which a central Tutor Agent (GPT-4o) interprets student input and orchestrates other components via a ReAct-style framework.
status review
generated by process:settle-candidates on 2026-10-10
trust unverified — no one has confirmed this page's content
sources 1 — jarosław-a-chudziak-and-adam-kostka-2025

Frontmatter is the page's machine-readable record — see the Schema & Guide.