Skip to main content
Hero Light

​
Transform how you learn

Sprout is an adaptive learning platform that creates personalized 3D knowledge graphs on any topic. Powered by seven specialized Claude AI agents, Sprout generates interactive learning pathways that adapt to your understanding in real-time. Built in 20 hours at a hackathon, Sprout reimagines AI-assisted learning by guiding you to discover answers yourself rather than simply providing them.

Quick Start

Get up and running with Sprout in under 10 minutes

Architecture Overview

Explore the multi-agent system and 3D knowledge graphs

API Reference

Complete REST API documentation for integration

Hand Tracking Setup

Navigate your knowledge graph with natural hand movements

​
Core Features

​
3D Knowledge Graphs

Visualize learning pathways as interactive three-dimensional networks. Each concept becomes a node you can explore, with edges representing prerequisite relationships. Navigate using your mouse, keyboard, or natural hand movements via OpenCV.

​
Seven Autonomous Agents

Sprout’s AI architecture orchestrates specialized Claude-powered agents that work together:
  • Topic Agent - Breaks topics into 6-10 prerequisite-linked concepts
  • Subconcept Bootstrap Agent - Generates 8-12 subconcepts and diagnostic questions per concept
  • Concept Refinement Agent - Personalizes learning paths based on diagnostic performance
  • Tutor Chat Agent - Teaches concepts chunk-by-chunk with interactive exercises
  • Grade Answers Agent - Provides detailed scoring and feedback
  • Generate Diagnostic Agent - Creates mixed-format assessments (MCQ and open-ended)
  • Review Learning Path Agent - Recommends enrichment and remediation after completion
All agents use a shared tool-calling loop that streams real-time progress via Server-Sent Events (SSE), giving you visibility into the AI’s reasoning process.

​
Adaptive Diagnostics

Before diving into a concept, Sprout assesses your current understanding with targeted diagnostic questions. Based on your answers, the Concept Refinement Agent restructures the subconcept graph to fill knowledge gaps or skip mastered material. 
// Example: Concept refinement flow from src/agents/concept-agent.ts
const tools = [
  { name: "grade_student_answers", ... },
  { name: "get_current_subconcepts", ... },
  { name: "check_student_history", ... },
  { name: "add_subconcept", ... },
  { name: "remove_subconcept", ... },
  { name: "add_prerequisite_concept", ... },
  { name: "validate_graph", ... }
];

​
Interactive Tutoring

The Tutor Agent breaks each subconcept into digestible chunks and guides you through with:
  • Multi-format exercises - Text, code, drawings, and quizzes
  • Worked examples - Step-by-step demonstrations
  • Mastery tracking - Progress persists across sessions
  • Voice support - Powered by ElevenLabs for audio guidance
Sprout follows one core methodology: the learner should reach the answer themselves. The AI is there to show the path, not give the solution.

​
Document Upload & Context

Upload PDFs, notes, or course materials when creating a topic. Sprout extracts relevant context and embeds it into concept nodes, ensuring the learning path aligns with your specific materials. 
# S3 document storage configured via backend/.env
AWS_ACCESS_KEY_ID=your_key
AWS_SECRET_ACCESS_KEY=your_secret  
AWS_REGION=us-east-1
AWS_S3_BUCKET=your-bucket

​
Technology Stack

Backend

  • Express 5 + TypeScript
  • Anthropic Claude SDK
  • Drizzle ORM + SQLite
  • AWS S3 for documents
  • Server-Sent Events (SSE)

Frontend

  • Next.js 16 + React 19
  • Three.js & React Flow
  • WebSocket hand tracking
  • Tailwind CSS v4
  • shadcn/ui components

Computer Vision

  • Python 3.11+
  • OpenCV + MediaPipe
  • WebSocket server (port 8765)
  • Real-time hand landmark detection

AI/ML

  • Anthropic Claude 3.5
  • Multi-agent orchestration
  • Tool-calling loops
  • Streaming responses

​
What Makes Sprout Different

Oxford University Press found that 80% of students rely on AI for schoolwork, and 62% say it’s making them worse at learning. AI tools are here to stay—the question is how students use them. Sprout doesn’t just give answers. It:
  1. Diagnoses understanding through adaptive assessments
  2. Personalizes structure by restructuring knowledge graphs based on your performance
  3. Guides discovery with Socratic tutoring that leads you to insights
  4. Tracks mastery across concepts to ensure retention
  5. Adapts in real-time as you learn, adding remediation or enrichment dynamically
Sprout is a single-user hackathon project with no authentication. All data is stored locally in SQLite. Do not use for production or multi-user scenarios without adding proper auth and database scaling.

​
Example Workflow

Here’s how a typical learning session works:
1

Create a topic

Describe what you want to learn: “Linear Algebra for Machine Learning” or “Fauna of Peru”
2

Upload materials (optional)

Add lecture notes, PDFs, or course documents to provide context
3

Watch agents build your graph

Stream real-time progress as agents generate 6-10 concepts and 8-12 subconcepts each
4

Take diagnostic assessments

Answer MCQ and open-ended questions to gauge your current understanding
5

Explore personalized paths

Navigate the adaptive 3D graph using mouse, keyboard, or hand tracking
6

Learn with the tutor

Dive into concepts with interactive exercises, worked examples, and voice guidance
7

Track your progress

View mastery scores and completed nodes across your knowledge network

​
Single-User Architecture

Sprout uses a hardcoded default user for simplicity: 
// src/index.ts - Auto-seeded on backend startup
const DEFAULT_USER_ID = "00000000-0000-0000-0000-000000000000";

async function ensureDefaultUser() {
  const existing = await db.select().from(users)
    .where(eq(users.id, DEFAULT_USER_ID));
  if (!existing.length) {
    await db.insert(users).values({
      id: DEFAULT_USER_ID,
      email: "[email protected]",
      title: "Default Learner",
    });
    console.log("Default user seeded.");
  }
}

​
Get Started

Ready to build your first knowledge graph? Head to the Quick Start guide to set up Sprout locally in under 10 minutes. For a deeper understanding of the agent architecture and workflows, check out the Architecture Overview.

Build docs developers (and LLMs) love

Get started for freeTalk to us