TutorCruncher has a built-in MCP server. MCP (Model Context Protocol) is the standard way AI assistants like Claude, ChatGPT and Codex connect to other systems. Once connected, you can ask your AI assistant to look things up and make changes in your account, for example, listing this week's Lessons, finding unpaid Invoices, adding a new Client, scheduling a Lesson, or recording a payment.
There are two ways to connect:
Claude and ChatGPT connect by signing in to TutorCruncher and approving the connection. No API key or settings file is needed.
Claude Code and Codex, which run in your terminal, connect using an API key.
Connecting Claude or ChatGPT
Note: To connect an AI assistant, your Administrator profile needs the 'Use and configure the API' permission. You can find out more about permissions here.
Step 1: Copy your connection address
In TutorCruncher, navigate to System > Settings > TutorCruncher API > MCP connections.
Click the copy button next to the address at the top of the page. For most accounts this will be
https://app.tutorcruncher.com/mcp.
This page also shows short setup steps for Claude and ChatGPT, and lists every assistant already connected to your Branch.
The MCP connections page, with the address to paste into your AI assistant.
Step 2: Add TutorCruncher to your AI assistant
Claude
These steps work on the Claude website and in the Claude desktop app.
In Claude, go to Customize > Connectors. In some versions of Claude this is under Settings > Connectors.
Click +, then Add custom connector.
Give it a name, such as 'TutorCruncher', and paste in the address you copied in Step 1.
Click Add, then Connect. Claude will open TutorCruncher so you can sign in and approve the connection (see Step 3).
Once connected, start a new conversation, click the + button at the bottom-left of the message box and select Connectors. Make sure TutorCruncher is switched on.
Note: If you're on a Claude Team or Enterprise plan, an Owner of your Claude organisation needs to add the connector first, from Organization settings > Connectors in Claude. Members can then click Connect on it from their own Connectors page. Claude's Free plan is limited to one custom connector.
ChatGPT
These steps work on the ChatGPT website. ChatGPT doesn't support custom connections in its mobile apps.
In ChatGPT, go to Settings > Apps > Advanced settings and turn on Developer mode. Older versions of ChatGPT call this area Connectors instead of Apps.
Go back to Settings > Apps and click Create.
Give it a name, paste in the address you copied in Step 1, and choose OAuth as the authentication method.
Click Create. If you see a Scan Tools button, click that first. ChatGPT will open TutorCruncher so you can sign in and approve the connection (see Step 3).
Once connected, start a new chat and select TutorCruncher from the tools menu, or mention it in your message.
Note: Which ChatGPT plans can do this is set by OpenAI. At the time of writing, full read and write access is available on ChatGPT Business, Enterprise and Edu plans. On Business, only workspace admins and owners can turn on Developer mode. On Pro, ChatGPT can read your TutorCruncher data but can't make changes. If you can't see the Developer mode toggle, check with whoever manages your ChatGPT workspace.
Step 3: Approve the connection
When your AI assistant sends you to TutorCruncher, sign in as normal. You'll then see a screen asking you to approve the connection. It shows:
The name of the application asking to connect, and the Branch it will be connected to.
The website your data will be sent to. Only continue if you recognise it and started the connection from there yourself.
A grid showing what the assistant will be able to reach: Clients, Tutors, Students, Jobs, Lessons and chats; Invoices, payments and Balances; Subjects, qualifications and Branch settings; and Safeguarding concerns.
Use the Read and Write tick boxes at the top of the grid to choose what the assistant can do. Untick Write if you only want it to look things up and never make changes. Write access always includes Read, so unticking Read unticks both.
Click Connect to approve, or Cancel to stop without connecting anything. You'll be sent back to Claude or ChatGPT, and TutorCruncher is ready to use.
Note: A connection can never do more than you can do yourself. It only reaches the Branch you're signed in to, and any area your own permissions don't cover is shown with a cross and stays off-limits. For example, Safeguarding/Wellbeing Concerns can only be reached by Administrators who are Safeguarding/Wellbeing Concern Managers.
Managing and disconnecting assistants
To see everything connected to your Branch, go to System > Settings > TutorCruncher API > MCP connections. Each connection shows the application, who connected it, what access it has, and when it was connected.
To end a connection, click Disconnect next to it and confirm. The assistant loses access straight away.
Connecting with an API key (Claude Code and Codex)
Claude Code and Codex run in your terminal and connect using an API key rather than by signing in.
Note: An API key gives access to your whole Branch, whatever the permissions of the Administrator who created it. Keep it private and treat it like a password. If you're using Claude or ChatGPT, use the sign-in method above instead.
Get your API key
In TutorCruncher, navigate to System > Settings > TutorCruncher API > API Integrations.
Select Add API Integration to add a new integration. You can call it 'TutorCruncher MCP'.
Adding an API Integration.
Once created, copy its Private Key.
The API Integration Private Key highlighted.
Claude Code
Run the command below, replacing YOUR_API_KEY with the Private Key you copied.
claude mcp add --transport http tutorcruncher https://app.tutorcruncher.com/mcp/ --header "Authorization: Bearer YOUR_API_KEY"
Codex
If you're using Codex, OpenAI's command line tool, run the command below to connect.
codex mcp add tutorcruncher --url https://app.tutorcruncher.com/mcp/ --bearer-token-env-var TC_API_KEY
Codex reads your key from an environment variable rather than from the command itself. Set it before starting Codex.
export TC_API_KEY="your-private-key-here"
To avoid setting the variable every session, add the export line to your ~/.zshrc on macOS, or your shell profile on Windows.
Verify the connection by running codex mcp list, or by running /mcp inside a Codex session.
What you can ask your AI assistants
"List all lessons scheduled for next week."
"Which invoices are unpaid?"
"Create a new client called Jane Smith with email jane@example.com."
"Show me this month's payments."
"Add a note to client 12345."
Frequently Asked Questions
Can I connect more than one Branch?
Yes, but each Branch is connected separately. A connection only reaches the Branch you were signed in to when you approved it. To connect another Branch, switch to it in TutorCruncher, then add the connector in your AI assistant again.
I set up Claude Desktop using the settings file. Do I need to change anything?
Your existing setup will keep working, as API keys are still accepted. We recommend switching to the sign-in method, though, as it lets you choose read-only access, limits the assistant to your own permissions, and can be disconnected from TutorCruncher at any time. It also no longer needs Node.js.
To switch, open Claude Desktop and go to Settings > Developer > Edit Config. Remove the "tutorcruncher" entry from claude_desktop_config.json, save the file and restart Claude Desktop. Then follow the Claude steps above.
Why can't my assistant see some areas of TutorCruncher?
When you connect by signing in, the assistant can only reach the areas your own Administrator permissions allow. If you can't see Invoices yourself, for example, neither can the assistant. The approval screen shows which areas are included before you connect. If you need access to more, ask an Administrator with the right permissions to update your profile. Changes to your permissions apply to your connection automatically.
I chose read-only access. How do I let the assistant make changes?
Go to System > Settings > TutorCruncher API > MCP connections, click Disconnect next to the connection, then connect again from your AI assistant and leave Write ticked on the approval screen.