Back

MCP server

Use PupilBot inside Claude, ChatGPT and Codex

PupilBot is an MCP server. Connect it once and your AI assistant can turn notes, articles and chapters into PupilBot studies, tell you when they are ready, and read back how well you know each topic.

The practising itself stays in the web app, the iPhone app or the Telegram bot. The assistant creates studies and reads your mastery; it never answers the questions for you.

Tools
4
Transport
Streamable HTTP
Auth
OAuth 2.1
Price
Free account

Server URL

https://www.pupilbot.com/mcp

Paste this into any MCP client. The first connection opens a PupilBot sign-in page where you approve the client — no API key to generate, nothing to configure. No account yet? Create one, it is free.

How to connect

Claude (web, desktop or mobile)

  1. Settings → Connectors → Add custom connector
  2. Name it PupilBot and paste https://www.pupilbot.com/mcp
  3. Claude opens the PupilBot sign-in — log in and approve
  4. Ask: “Turn these lecture notes into a PupilBot study.”

Claude Code

Run this in your terminal:

claude mcp add --transport http pupilbot https://www.pupilbot.com/mcp

The first call opens the PupilBot sign-in in your browser. Approve it once and the connection persists.

Codex CLI

Add the server, then sign in — adding an OAuth-protected server does not start the sign-in on its own:

codex mcp add pupilbot --url https://www.pupilbot.com/mcp
codex mcp login pupilbot

ChatGPT

  1. Settings → Apps & Connectors, and turn on Developer mode if you do not see custom connectors
  2. Create a connector with the URL https://www.pupilbot.com/mcp and keep OAuth as the authentication
  3. Connect, sign in to PupilBot and approve
  4. Ask: “Which topics am I weakest on across all my studies?”

The menu path moves between ChatGPT releases; OpenAI’s connect guide has the current steps, and which plans expose it is ChatGPT’s call, not ours.

Cursor, Zed and other MCP clients

  1. Add an MCP server with transport http (streamable HTTP) and the URL https://www.pupilbot.com/mcp
  2. Complete the sign-in the client opens — clients register themselves, there is nothing to provision on our side

What your assistant can do

4 tools, all scoped to your own account. It can create studies and read your progress; it cannot delete anything or see another student’s data.

ToolWhat it does
create_study Turns text your assistant hands it (notes, an article, a chapter) into a new study with topics and practice questions. Takes about a minute.
get_study_status Whether a study is still processing, ready to practise, or failed.
list_studies Your studies, newest first.
get_mastery How well you know things: a summary across every study, or per-topic scores, streaks and next review dates for one study.

Things to ask

“Turn these lecture notes into a PupilBot study.”
“Make a study from this article and tell me when it is ready.”
“Which topics am I weakest on across all my studies?”
“List my studies and tell me what is due for review today.”

Privacy and access

For machines

Streamable HTTP, stateless: one JSON-RPC 2.0 request per POST, one JSON response. No SSE stream, no session ids, no JSON-RPC batching. A GET that asks for text/event-stream, and any DELETE, answer 405 — which is what tells a client to fall back to plain POSTs; a GET that does not ask for the stream gets this page. Protocol versions: 2025-06-18, 2025-03-26, 2024-11-05.

Discovery: /.well-known/oauth-authorization-server (RFC 8414) and /.well-known/oauth-protected-resource (RFC 9728). An unauthenticated POST answers 401 with a WWW-Authenticate challenge naming the second one.