
TutorFlow Agent Platform is an AI-powered education API that lets developers integrate automated grading, answer evaluation, and course generation into any product. One API call evaluates a student answer and returns structured, scored feedback. Another generates a complete interactive course from a single prompt.
It works with any language or framework, connects directly to MCP-compatible AI agents like Claude and GPT, and bills a fixed number of credits per call with no token math. If you're building an EdTech product and need reliable AI assessment infrastructure, this is what TutorFlow was built for.
Why Building an AI Grading System from Scratch Is Costly
If you're an EdTech developer, you've probably considered plugging an LLM directly into your grading pipeline. It sounds simple until you actually try it.
Getting consistent, rubric-aligned scoring across subjects and languages is not a weekend project. You need prompt chains that produce reliable structured output, not raw text you can't pipe into a gradebook. You need normalized scores, typed feedback fields, and schemas your frontend can trust.
Then there's model orchestration. Which provider gives the best results for chemistry vs. essay grading? What happens when a model degrades or a provider has downtime? You need fallback logic, cost optimization, and the ability to switch providers without rewriting your integration.
And once you ship, the work continues. Models change, evaluation quality drifts, new languages get requested, and edge cases keep appearing. These are real engineering costs that distract from building your actual product.
TutorFlow Agent Platform absorbs all of it so your team can focus on the learning experience, not the infrastructure behind it.
What Does TutorFlow Agent Platform Include?

TutorFlow Agent Platform gives developers a complete education API layer. Content work runs on the Content API in a classroom of your account, and the Agent Platform handles the account, keys, and billing:
- AI Evaluation API: Grade open-ended answers, rubric-based assessments, and exact-match questions with structured AI feedback.
- Course Generation API: Create full interactive courses with chapters, lessons, quizzes, and coding exercises from a single prompt.
- More generators: Tests, single-lesson modules, slide decks, games, simulations, and narrated videos through the same API.
- MCP Server: MCP-compatible AI agents (Claude, GPT, Gemini) connect directly to TutorFlow tools. No HTTP boilerplate needed.
- Agent Self-Service: Autonomous agents can register an account, get 50 trial credits, and start working without human intervention.
- Webhooks: A signed request when a generation run, a video render, or an evaluation finishes, instead of polling.
- Multilingual Feedback: AI-generated feedback in 40+ languages.
- Credit Pricing: Every priced call charges a fixed number of credits at 2 cents each. An evaluation is 1 credit. No subscriptions, no token math.
How to Evaluate a Student Answer with One API Call
Register an account first. One call returns a Content API key and the id of your default classroom:
curl -X POST https://api.tutorflow.io/v1/platform/agent/register \
-H "Content-Type: application/json" \
-d '{"name":"My Agent","slug":"my-agent","email":"you@example.com"}'Keep contentApiKey as TUTORFLOW_CONTENT_API_KEY and defaultClassroomId as CLASSROOM_ID. Then send a student's answer to the evaluation endpoint. You get back a structured response with a score, strengths, suggestions, and a summary.
curl -X POST "https://api.tutorflow.io/v1/content/classrooms/$CLASSROOM_ID/evaluations" \
-H "Authorization: Bearer $TUTORFLOW_CONTENT_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"evaluationType": "open_ended",
"questionText": "What causes seasons on Earth?",
"learnerAnswer": "The tilt of Earth axis causes different amounts of sunlight to reach each hemisphere throughout the year.",
"language": "en",
"maxScore": 10,
"learnerExternalId": "learner-42"
}'The response gives you everything you need to display feedback or store in your gradebook. This one is shortened:
{
"id": "2f4ad455-1e8e-4b6c-9f3a-7ebf4cf6f483",
"status": "COMPLETED",
"evaluationType": "open_ended",
"learner": { "externalId": "learner-42", "name": null },
"score": 8,
"maxScore": 10,
"normalizedScore": 80,
"strengths": [
"Correctly identifies axial tilt as the primary cause",
"Connects tilt to variation in sunlight across hemispheres"
],
"suggestions": [
"Mention how the angle of sunlight affects energy distribution",
"Explain the relationship between tilt and seasonal daylight duration"
],
"feedbackSummary": "Strong understanding of the core mechanism behind seasons...",
"createdAt": "2026-04-04T12:00:00.000Z",
"completedAt": "2026-04-04T12:00:01.000Z"
}TutorFlow supports three evaluation types: open_ended, rubric_short_answer with a rubric, and exact against a reference answer. Each evaluation costs 1 credit ($0.02). learnerExternalId is your own id for the learner, and it builds a per-learner mastery model you can read back.
How to Generate an AI-Powered Course with One API Call
The Course Generation API creates a complete curriculum with chapters, interactive lessons, and coding exercises. The AI selects lesson types based on the topic: coding tutorials, quizzes, flashcards, AI tutor conversations, and more.
curl -X POST "https://api.tutorflow.io/v1/content/classrooms/$CLASSROOM_ID/courses/generate" \
-H "Authorization: Bearer $TUTORFLOW_CONTENT_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: course:intro-python:v1" \
-d '{
"topic": "A beginner Python course covering variables, loops, and functions with hands-on exercises",
"level": "beginner",
"lessonCount": 6,
"language": "en"
}'TutorFlow writes the course in the background and answers 202 Accepted with a run to follow:
{
"runId": "6bff...",
"status": "queued",
"kind": "course",
"estimatedCredits": 21,
"statusUrl": "/v1/content/classrooms/0fce.../runs/6bff..."
}Read statusUrl until status is completed, or register a webhook and get told. resourceId on the run is the course id. Reading the course gives you two links:
{
"id": "5d2a...",
"title": "Python Quick Start",
"level": "beginner",
"visibility": "PUBLIC",
"editorUrl": "https://tutorflow.io/org/my-agent/my-agent/courses/chat/09cf.../lessons",
"publicUrl": "https://my-agent.tutorflow.ac/courses/python-quick-start"
}editorUrl opens the course in the TutorFlow editor for the person who reviews it. publicUrl is the course page learners open. A course costs 3 credits for the outline plus 6 per lesson, so this six-lesson course is 39 credits ($0.78).
How to Connect AI Agents via Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an open standard that lets AI agents discover and call tools directly. TutorFlow provides a hosted MCP server, so agents like Claude Code can integrate without writing HTTP calls.
Add TutorFlow to your .mcp.json:
{
"mcpServers": {
"tutorflow": {
"type": "http",
"url": "https://mcp.tutorflow.io/api/mcp",
"headers": {
"Authorization": "Bearer tf_content_..."
}
}
}
}Or connect programmatically from any MCP client:
import { Client } from '@modelcontextprotocol/sdk/client/index.js';
import { StreamableHTTPClientTransport } from '@modelcontextprotocol/sdk/client/streamableHttp.js';
const transport = new StreamableHTTPClientTransport(
new URL('https://mcp.tutorflow.io/api/mcp'),
{
requestInit: {
headers: { Authorization: `Bearer ${process.env.TUTORFLOW_CONTENT_API_KEY}` },
},
},
);
const client = new Client({ name: 'my-agent', version: '1.0.0' });
await client.connect(transport);Once connected, your agent gets access to structured tools:
| Tool | Description |
|---|---|
register_organization | Create an account and receive a Content API key and trial credits |
create_course | Start a course generation run |
get_course_generation | Follow a run and read the finished course |
list_courses | List courses with their editor links |
run_evaluation | Grade an answer and return the score and feedback |
list_evaluations | List past evaluations with filtering |
get_pricing | The live price list |
get_account | Account info and credit balance |
check_billing_status | Payment method, balance, and billing mode |
create_billing_session | A checkout link to send to a person for buying credits |
Agents can also self-onboard entirely through MCP: register an account, have a person verify the email, set up billing, and start evaluating, without anyone touching the dashboard.
Production-Ready Infrastructure
TutorFlow Agent Platform is designed to run in production, not just demos:
- Scoped API keys with secure hashing, rotation, and separate test keys that spend no credits
- Signed webhooks when a generation run, a video render, or an evaluation finishes
- Idempotency keys so a retried request never starts or charges twice
- Background runs with status, progress, and credits charged so far
- Multi-model backend that switches providers without changing your integration code
- Rate limits reported in response headers, so your client can pace itself
- One credit balance shared by API calls and the TutorFlow editor, with a full credit history
Pricing: Fixed Credits per Call, No Subscriptions
No subscriptions, no token math, no surprises. Every priced call charges a fixed number of credits, and a credit costs 2 cents. Reads and updates are free.
| Call | Credits | USD |
|---|---|---|
| Evaluation | 1 | $0.02 |
| Course | 3, plus 6 per lesson | $0.06, plus $0.12 per lesson |
| Test | 3 | $0.06 |
| Single-lesson module | 6 | $0.12 |
A new account starts with 50 trial credits. After that, once the account's email is verified, buy credits by card, or with USDC where x402 is enabled, starting at $5, or add a card for metered use. The full price list is in Pricing and Billing. For high-volume teams, enterprise contracts with an SLA or invoiced billing are available.
Get Started in 3 Steps
1. Register an account with one API call, the register_organization MCP tool, or the TutorFlow CLI. You get a Content API key, a classroom, and 50 trial credits.
2. Make your first API call. Evaluate an answer or generate a course using the examples above.
3. Go deeper. Connect your AI agent via MCP, set up webhooks, or follow a generation run end to end.
Explore the full documentation at tutorflow.io/agent-platform/docs, or jump straight to the Quick Start guide.
Frequently Asked Questions
What is TutorFlow Agent Platform?
TutorFlow Agent Platform is the account layer for developers and AI agents that use TutorFlow: registration, API keys, and credit billing. Content work, such as AI grading, answer evaluation, and course generation, runs on the TutorFlow Content API. It supports open-ended, rubric-based, and exact-match evaluation types with structured feedback in 40+ languages.
How is TutorFlow different from calling an LLM directly?
Calling an LLM directly requires you to build and maintain prompt engineering, output parsing, scoring normalization, model fallback logic, and multilingual support yourself. TutorFlow handles all of that behind a single API endpoint and returns typed, structured responses you can use directly in your product.
What evaluation types does TutorFlow support?
TutorFlow supports three evaluation types: open-ended (essay-style answers graded with AI feedback), rubric-based (scored against structured criteria), and exact-match (checked against a reference answer). Each evaluation costs 1 credit.
Does TutorFlow work with AI agents like Claude or GPT?
Yes. TutorFlow provides a hosted MCP (Model Context Protocol) server that lets MCP-compatible AI agents connect directly to TutorFlow tools. Agents can register an account, set up billing, generate courses, and grade answers without any human setup.
How much does TutorFlow Agent Platform cost?
A credit costs 2 cents. An evaluation is 1 credit ($0.02), and a course is 3 credits plus 6 per lesson. There are no subscriptions. A new account gets 50 trial credits, and you buy more starting at $5 and pay only for what you use.


