The Model Context Protocol (MCP for short) connects teamspace to an AI assistant of your choice – Claude, ChatGPT, Cursor or VS Code, for example. The assistant can then work with teamspace on your behalf: analyse data, explain how things connect, create open items. You ask in plain language, and the assistant fetches the answer from your tenant.
Two ways to use AI with teamspace
Before you read on, here is a distinction that saves a lot of confusion – because both are casually called “AI in teamspace”, yet the two are fundamentally different:
| MCP (this topic) | AI features | |
|---|---|---|
| What happens? | Your assistant reaches into teamspace from the outside | teamspace uses an AI internally |
| Where do you work? | In Claude, ChatGPT, Cursor … | In teamspace itself |
| Who sets it up? | An admin defines the MCP, you connect your client | An admin connects an AI and switches features on |
| What do you do for it? | Connect once and authorise via OAuth | Nothing – the feature is simply there |
| Example | ”Show me all open bug tickets grouped by status” | Have a ticket summarised at the click of a button |
So when you connect your ChatGPT account to teamspace via MCP, that is not AI used in teamspace – it is your assistant using teamspace as a tool. Conversely, the ticket summary inside teamspace needs no client at all and no MCP.
The two do not rule each other out, and many companies use both. How the built-in features work is covered in Introduction to AI features.
What you get out of it
The appeal of MCP is that you don’t have to decide in advance what you want to ask. A report in teamspace answers the question it was built for. An assistant with MCP access answers whatever question comes to mind:
- Analyse and condense – “Which projects have used more hours than planned this month?”
- Capture without switching windows – “Create an open item for Anna on that, due Friday.”
- Find connections – link tickets, projects and contacts in a single question, without opening three modules.
- Move data – right up to reconciling two systems or running a migration.
The limits here are set not by the assistant but by your permission profile: via MCP you see and change exactly what you would be allowed to in teamspace itself – not a single record more. Why that is, and what follows from it, is explained in How MCP works in teamspace.
How the pieces fit together
Three building blocks, no more:
- The MCP configuration – an administrator clicks together, in teamspace, which tools an MCP offers, and releases it to particular user groups. Each configuration has an address of its own.
- Your client – Claude, ChatGPT, Cursor or another assistant, in which you enter that address as a connector.
- The authorisation via OAuth – you log in once and confirm what the client may do. From then on it works under your name and with your permissions.
By design there is not one big MCP that can do everything, but as many tailored ones as you like. A general MCP with read-only access to non-critical data can be open to everyone, whereas an HR MCP with broader capabilities is open only to the HR department.
Where to start
- As an administrator: Create an MCP configuration, then Select and tailor tools.
- As a user: Connect teamspace to Claude – the process is the same with other clients, only the menus have different names. Which clients are tested is listed in Supported clients.
Availability: MCP is included in all editions – unlike the plain REST API, which is available from the enterprise edition upwards. From September 2026 the integration is enabled for all tenants; interested parties and beta customers can already test it on request.