AI Tutoring

Build AI tutoring platforms where each student gets their own persistent environment — learning state, progress, exercises, and conversation history that persists across every session.

The problem

Personalized AI tutoring needs per-student state. Each student has different knowledge gaps, different progress, different exercises. You need to track what they've learned, what they've struggled with, and adapt. Storing this in a shared database works until you need code execution, file storage, or search — then you're stitching together services.

How OnCell solves it

Each student gets their own cell. The tutor agent stores learning progress in the database, exercises as files, and indexes past interactions for context. When a student returns days later, everything is exactly where they left off.

const cell = await oncell.cells.create({
  customerId: "student-789",
  agent: tutorAgentCode,
});

// Inside the agent:
const progress = ctx.db.get("progress") || { level: 1, topics: [] };
const pastInteractions = ctx.search.query(question);
// Adapt response based on student's history
ctx.db.set("progress", { ...progress, lastTopic: "algebra" });

What your agent gets

Per-student state — learning progress, skill levels, exercise history. Isolated and persistent.

Conversation memory — full history of every interaction. Search across past conversations to maintain context.

Code execution — for coding tutors: students write code, agent runs it, checks output, gives feedback. All in their own sandbox.

Pause / resume— student leaves, cell pauses. Student returns a week later, resumes in 200ms. You pay almost nothing while they're away.

Example architectures

Coding tutor — student writes code in their cell, agent runs tests, gives hints. Progress tracked per student.

Language tutor — tracks vocabulary, grammar mistakes, speaking patterns. Adapts difficulty based on history.

Math tutor — generates personalized exercises, tracks which concepts are mastered, adjusts curriculum.