About three.ws Tutor
three.ws Tutor is an MCP server for AI agents that need to inspect and finalize Pay-As-You-Learn tutoring sessions. It returns an itemized tab and seals the final invoice with a SHA-256 attestation; both operations are free.
Key features
- 01Read itemized session tabs
- 02View question costs and running total
- 03Close sessions with attested invoices
- 04Repeat close without changing the invoice
Use cases
- 01Check charges during a tutoring session
- 02Finalize a session and obtain its invoice
MCP support
Exposes `load_session` and `close_session` tools to MCP clients over stdio. Sessions are identified by `sessionId`; the tools do not require an API key or signer.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @three-ws/tutor-mcpTHREE_WS_BASEBase URL of the three.ws API that serves the tutor session endpoint. Override only when self-hosting or targeting a preview deployment.THREE_WS_TIMEOUT_MSPer-request timeout in milliseconds for the tutor session reads and the close call.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
2Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| sessionId | string | Yes | Opaque tutoring session identifier (the id the session was opened with). Truncated to 100 chars by the API. |
Details
From the project README@three-ws/tutor-mcp
Manage a Pay-As-You-Learn tutoring session from any AI agent — read its itemized running tab and close it for an attested invoice.
A Model Context Protocol server that exposes the three.ws Pay-As-You-Learn tutor ledger over stdio. Read a tutoring session's itemized tab — one line item per answered question, with per-item cost and a running total — and close the session to seal an attested invoice.
The tutor bills per answered question; this server's tools are free. Viewing the running tab and closing the session never charge the learner — a learner must never pay to see what they owe or to end the session. Sessions are addressed by an opaque sessionId, so no API key and no signer are required — point THREE_WS_BASE at a deployment and go.
Install
npm install @three-ws/tutor-mcp
Or run with npx (no install):
npx @three-ws/tutor-mcp
Quick start
Claude Code, one line:
claude mcp add tutor -- npx -y @three-ws/tutor-mcp
Claude Desktop / Cursor (claude_desktop_config.json or mcp.json):
{
"mcpServers": {
"tutor": {
"command": "npx",
"args": ["-y", "@three-ws/tutor-mcp"]
}
}
}
Inspect the surface with the MCP Inspector:
npx -y @modelcontextprotocol/inspector npx @three-ws/tutor-mcp
Tools
| Tool | Type | What it does |
|---|---|---|
load_session | read | Read a session's itemized running tab by sessionId — line items (question, level, tokens, cost), status, and running total. |
close_session | write | Finalize a session and return its attested invoice — seals the ledger with a SHA-256 attestation and locks further billing. |
Both tools are free and move no funds: the per-question answers are the paid action (handled elsewhere). load_session reads live data — the tab grows as questions are billed, so it is not idempotent. close_session mutates state (open → closed) but is itself idempotent: re-closing returns the same itemized total.
Input parameters
load_session — sessionId (required, ≤100 chars).
close_session — sessionId (required, ≤100 chars).
Example
// load_session
> { "sessionId": "learn-abc123" }
{
"ok": true,
"sessionId": "learn-abc123",
"status": "open",
"createdAt": "2026-06-24T03:40:00.000Z",
"questionCount": 2,
"lineItems": [
{ "n": 1, "question": "What is a Solana PDA?", "level": "intro", "outputTokens": 180, "costAtomics": 10000, "costUsd": "0.010000", "at": "2026-06-24T03:41:00.000Z" },
{ "n": 2, "question": "How do CPIs work?", "level": "intro", "outputTokens": 220, "costAtomics": 10000, "costUsd": "0.010000", "at": "2026-06-24T03:42:00.000Z" }
],
"totalAtomics": 20000,
"totalUsd": "0.020000"
}
// close_session
> { "sessionId": "learn-abc123" }
{
"ok": true,
"sessionId": "learn-abc123",
"createdAt": "2026-06-24T03:40:00.000Z",
"closedAt": "2026-06-24T03:45:00.000Z",
"questionCount": 2,
"lineItems": [ /* same itemized line items */ ],
"totalAtomics": 20000,
"totalUsd": "0.020000",
"attestation": "sha256:…"
}
A sessionId with no stored history returns an empty open session (questionCount: 0) rather than an error — an honest "nothing billed yet", not a failure.
Requirements
- Node.js >= 20.
- Network access to
https://three.ws(or your ownTHREE_WS_BASE).
Environment variables
| Variable | Required | Default |
|---|---|---|
THREE_WS_BASE | no | https://three.ws |
THREE_WS_TIMEOUT_MS | no | 20000 |
Links
- Homepage: https://three.ws
- Changelog: https://three.ws/changelog
- Issues: https://github.com/nirholas/three.ws/issues
- License: Apache-2.0, see LICENSE
Part of the three.ws SDK suite — 3D AI agents, on-chain identity, and agent payments.
Website · Changelog · GitHub
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.