For AI agents and the people who build them

TutorFlow Agent Platform

An API and MCP server your agent uses to build courses and grade answers
It signs itself up and pays per action.

API Call

<3s

Avg Evaluation

40+

Languages

Feedback Fields

Get started

Your agent's first four calls

Open one step at a time. Step 1 sets TUTORFLOW_KEY and CLASSROOM_ID, which the other steps use.

Needs curl and jq. Change the slug and email before you run it.

1Register the agent

One call with no key. The answer holds contentApiKey, defaultClassroomId, and 50 trial credits. The key is shown once, so keep it.

bash

RESPONSE=$(curl -s -X POST https://api.tutorflow.io/v1/platform/agent/register \
  -H "Content-Type: application/json" \
  -d '{"name":"My Agent","slug":"my-agent-2048","email":"you@example.com"}')
export TUTORFLOW_KEY=$(echo "$RESPONSE" | jq -r .contentApiKey)
export CLASSROOM_ID=$(echo "$RESPONSE" | jq -r .defaultClassroomId)

2Start a course

TutorFlow writes the course in the background and answers 202 with a runId. A one-lesson course costs 6 credits.

bash

RUN_ID=$(curl -s -X POST \
  "https://api.tutorflow.io/v1/content/classrooms/$CLASSROOM_ID/courses/generate" \
  -H "Authorization: Bearer $TUTORFLOW_KEY" \
  -H "Content-Type: application/json" \
  -d '{"topic":"Fractions for grade 4","lessonCount":1}' | jq -r .runId)

3Follow the run

Read the run every 20 to 30 seconds until status is completed or failed. When it completes, resourceId holds the new course's ID.

bash

curl -s "https://api.tutorflow.io/v1/content/classrooms/$CLASSROOM_ID/runs/$RUN_ID" \
  -H "Authorization: Bearer $TUTORFLOW_KEY" | jq '{status, resourceId, error}'

4Grade an answer

Evaluations answer at once with a score, strengths, and suggestions. Each one costs 1 credit.

Request

bash

curl -s -X POST "https://api.tutorflow.io/v1/content/classrooms/$CLASSROOM_ID/evaluations" \
  -H "Authorization: Bearer $TUTORFLOW_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "evaluationType": "open_ended",
    "questionText": "Explain photosynthesis",
    "learnerAnswer": "Plants use sunlight to convert CO2 and water into glucose.",
    "maxScore": 10,
    "language": "en",
    "learnerExternalId": "learner-42"
  }'

Response

json

{
  "id": "5023...",
  "status": "COMPLETED",
  "evaluationType": "open_ended",
  "learner": { "externalId": "learner-42", "name": null },
  "score": 7,
  "maxScore": 10,
  "normalizedScore": 70,
  "strengths": ["Correctly identifies sunlight as key factor"],
  "suggestions": ["Mention the role of chlorophyll", "Explain the chemical equation"],
  "feedbackSummary": "Good foundational understanding. To improve, include the role of chlorophyll."
}

Prefer a terminal? Use the CLI

@tutorflow/cli makes the same calls. agent register saves the key to ~/.tutorflow/config.json, so the commands after it need no key or classroom flag.

bash

npm install -g @tutorflow/cli
tutorflow agent register --name "My Agent" --slug my-agent-2048 --email you@example.com
tutorflow course generate "Fractions for grade 4" --lessons 1 --wait
tutorflow evaluate --type open-ended --question "Explain photosynthesis" \
  --student-answer "Plants use sunlight to make glucose." --learner-id learner-42

Read the CLI guide

What your agent can do

What your agent can do without asking anyone

One Content API key works over REST, MCP, and the CLI. REST reaches everything; MCP covers sign-up, billing, courses, and grading.

Grade learner answers

Send an answer with a rubric or a reference answer. The score, strengths, and suggestions come back as typed fields.

Make teaching material

Start a course, test, slide deck, video, or module with one call, then follow the run until it finishes.

Call it as MCP tools

Add the server URL and a key, and Claude Code or any MCP client can use TutorFlow as tools.

Sign up without a person

Your agent registers its own account and gets a key, a classroom, and 50 trial credits in the same answer.

Know what each call costs

Every action draws a fixed number of credits, at 2 cents each, from one balance, and estimatedCredits arrives before the run starts. Rate limits come back in response headers.

Keys for each job

Keep test and live keys apart, scope each key, and rotate it. A test key works in a sandbox classroom and spends nothing.

Hear when work finishes

In the language your agent asks for

Content and feedback come back in the language your agent asks for.

MCP Server

Built for agents, not just humans.

Your agent finds TutorFlow's tools on its own: sign up, check the balance, generate a course, follow the run, and grade answers, with no HTTP code to write.

Add one server entry to your MCP client's config, and Claude, Gemini, or any other MCP client connects
Your agent signs itself up and gets a key in the same call
When credits run low, it sends a person a checkout link or, where x402 is enabled, pays in USDC itself
Typed inputs, so a bad call comes back with the reason

MCP guide Agent onboarding

Claude Code

json

// .mcp.json: add to your project root
{
  "mcpServers": {
    "tutorflow": {
      "type": "http",
      "url": "https://mcp.tutorflow.io/api/mcp",
      "headers": {
        "Authorization": "Bearer tf_content_..."
      }
    }
  }
}

MCP SDK

typescript

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 tf_content_...' },
    },
  },
)

const client = new Client({ name: 'my-agent', version: '1.0.0' })
await client.connect(transport)

// Call tools directly
const result = await client.callTool('run_evaluation', {
  evaluationType: 'open_ended',
  questionText: 'Explain photosynthesis',
  learnerAnswer: 'Plants use sunlight to make glucose...',
})

Self sign-up

json

// No auth needed: agents register themselves
POST https://api.tutorflow.io/v1/platform/agent/register

{
  "name": "My AI Tutor",
  "slug": "my-ai-tutor",
  "email": "agent@example.com",
  "agentIdentity": "claude-3-5-sonnet"
}

// Returns a Content API key and the classroom it works in
// { "contentApiKey": "tf_content_...", "defaultClassroomId": "...", "trialContentCredits": 50 }

Pricing

Pay for what your agent makes

Every call draws from one credit balance on your account, at 2 cents a credit. This is the live price list your agent is charged from.

What you make	Credits	US dollars
A course with its outline and every lesson	3 credits plus 6 credits per lesson	$0.06 plus $0.12 per lesson
One lesson module	6 credits per lesson	$0.12 per lesson
A test with auto-graded questions	3 credits per test	$0.06 per test
A text page in a slide deck	1 credit per text page	$0.02 per text page
A slide page with an AI image	3 credits per image page	$0.06 per image page
Narrating a video's scenes	1 credit per 200 narration characters	$0.02 per 200 narration characters
Rendering a video to MP4	3 credits per render	$0.06 per render
A game ready to play	41 credits per game	$0.82 per game
Grading one answer	1 credit per request	$0.02 per request

How billing works

Top up from $5 by card, or with USDC on Base, once your email is verified.
With a card on file, work keeps going after your credits run out, and that use is billed to the card.
A new account starts with 50 free credits.
Test keys work in a sandbox classroom and spend no credits.

Read the pricing guide

Ready when your agent is

TutorFlow Agent Platform

Your agent's first four calls

What your agent can do without asking anyone

Grade learner answers

Make teaching material

Call it as MCP tools

Sign up without a person

Know what each call costs

Keys for each job

Hear when work finishes

In the language your agent asks for

Built for agents, not just humans.

Pay for what your agent makes

Let your agent make its first course today