Cezanne Learning MCP Connector (alpha testing)
The Cezanne Learning MCP Connector connects Cezanne Learning to Claude, Anthropic's AI assistant, allowing you to ask questions in plain English and get answers using live data from your Cezanne Learning instance.
MCP (Model Context Protocol) is an open standard that allows Claude to connect to external systems. Once connected, Claude can access data available through the Cezanne Learning API, including users, teams, learning content, enrolments and activities.
The connector is currently in Early Access. Access is enabled by your Cezanne Learning account team, and some parts of the experience are still being developed.
What you can ask
Once you're connected, you can explore your Cezanne Learning data conversationally. For example:
- “Which learners haven't started their onboarding pathway?”
- “Give me a completion summary for the Leadership Essentials course.”
- “Which channels have had the most content added in the last 30 days?”
- “What learning content do we have tagged under 'compliance'?”
Claude reads your data using your access token, so it sees exactly what your own account can see.
Availability and what's coming
We're rolling the connector out gradually so we can support early adopters closely. Here's an honest picture of where things stand.
What's here now
- Connect Claude Desktop or Claude Code to your Cezanne Learning instance and query your data live.
- Your Cezanne Learning account team enables access per company, you won't be able to connect until they've switched it on for you.
- Your account team gives you the exact server address to use, so there’s nothing to work out about regions or hosting.
- Mint your own API access token from Settings then Company then API, any Owner or Admin can generate one without involving your account team.
Keep in mind
Your token currently has full access. During Early Access, the access token carries the same authority as your admin account. It can read, create, change and delete data through the API. Treat it like an admin password: don’t share it, don’t post it in chat tools, and only paste it into Claude clients you trust. If in doubt, regenerate it (which instantly revokes the old one).
What's coming next
- Read-only access: scoped tokens that can look but not touch, so you can safely hand Claude read access without write or delete authority.
- One-click sign-in: signing in to Cezanne Learning from Claude directly (including from claude.ai in the browser), instead of generating and pasting a token.
- Listing in Claude's connector directory: so connecting Cezanne Learning becomes a built-in option inside Claude.
Pre-requisites
Before you connect, you'll need three things:
- Early Access switched on for your company. Contact your Cezanne Learning account team to request it, they enable the connector and give you your server address.
- API access in your Cezanne Learning package. If you already use the Cezanne Learning API, you have this.
- A Claude client that supports remote MCP servers, such as Claude Desktop or Claude Code. Connecting from claude.ai in the browser isn't supported yet.
Role requirements
| Action | Roles |
| Generate an access token | Owner, Admin |
| Use the connection (what Claude can see) | Whatever your own account can access |
Quick Start Guide
Step 1: Ask your account team to enable Early Access
Your Cezanne Learning account team switches the connector on for your company and gives you your server address. You can't connect until this is done, so this is always the first step.
Step 2: Mint your own access token
You generate the token yourself, there's no need to ask your account team for one. Any Owner or Admin can mint it from inside Cezanne Learning:
- Log in to Cezanne Learning as an Owner or Admin
- From the sidebar, go to Settings then Company then API (labelled API & sFTP if your plan includes SFTP transfers)
- In the MCP access token section, select Generate MCP access token
- Copy it straight away, for security, it won't be shown again. If you lose it, generate a new one
Tokens are valid for 30 days. Generating a new token revokes your previous one, so you always have exactly one live token, and regenerating is how you revoke a token you no longer trust.
Step 3: Connect Claude
Use the server address your account team gave you.
Claude Code
claude mcp add --transport http cezannelearning \ YOUR-SERVER-ADDRESS \ --header "Authorization: Bearer YOUR-TOKEN-HERE"
Claude Desktop (and other apps that use a config file)
Claude Desktop connects through a small bridge utility called mcp-remote, which needs Node.js installed. Go to Settings then Developer then Edit Config and add:
{ "mcpServers": { "cezannelearning": { "command": "npx", "args": [ "mcp-remote", "YOUR-SERVER-ADDRESS", "--header", "Authorization: Bearer YOUR-TOKEN-HERE" ] } }}
Save and restart Claude Desktop. Cezanne Learning appears in the tools menu once connected.
💡Tip
Your server address ends in /mcp. If your account team gave you an address ending in /sse, ask them for the updated one, the older format is no longer supported.
Step 4: Test it
Ask Claude something simple, such as “Can you list the first 10 users in Cezanne Learning?” If Claude answers with real data from your platform, you’re connected.
FAQs
How do I get access?
Contact your Cezanne Learning account team. Early Access is enabled per company, and they'll switch it on and give you your server address.
Once that's done, you can mint your own access token — you don't need to go back to them for it.
How is this different from Volt?
Volt is Cezanne Learning's built-in AI assistant, ready to use inside the platform.
The MCP Connector works the other way round: it lets you bring your Cezanne Learning data into your own external assistant, such as Claude.
Use Volt for help inside Cezanne Learning; use the connector when you want your Cezanne Learning data in the AI tools your team already works in.
Do I need to know which region or pod my platform is on?
No. Your account team gives you the exact server address to use, so there's nothing to look up.
How long does my token last?
30 days.
When Claude starts reporting that it can't connect, generate a fresh token from Settings > Company > API and update it in your Claude configuration.
What can Claude actually do with my data?
Claude sees and acts with the same authority as your account.
During Early Access, every token carries full read, write, and delete access, so treat it like an admin credential and only connect clients you trust.
Scoped read-only access is coming.
Can I connect from claude.ai in the browser?
Not yet. Browser-based connectors need a sign-in flow that's on the roadmap.
For now, use Claude Desktop or Claude Code.
Is my data shared with anyone?
Claude accesses your Cezanne Learning data using your token and handles it under your organisation's agreement with Anthropic.
Anything Claude retrieves becomes part of your conversation with Claude, so check your own AI and data-handling policies before connecting.
Troubleshooting
| Issue | Solution |
| Claude says it can't connect, or authorisation failed. | The token is wrong, expired or revoked. Generate a fresh one from Settings then Company then API and update your configuration. |
| “Access denied” when connecting. | Early Access isn't enabled for your company yet, or your package doesn't include API access. Contact your Cezanne Learning account team. |
| The MCP access token section is missing in Settings. | Either the connector isn't enabled for your company yet, or you're not signed in as an Owner or Admin. |
| Claude connects but answers are slow or requests fail now and then. | Fair-use rate limits keep the service responsive for everyone. We may briefly throttle heavy automated use, try again shortly. |
Data and privacy
Claude accesses Cezanne Learning with your token, so it sees what your account can see. Anything Claude retrieves becomes part of your conversation with Claude and is handled under your organisation’s agreement with Anthropic. Check your organisation’s AI and data-handling policies before connecting, and make sure the right people know the integration is in place.