Connecting an AI Assistant to ContactDrive

Connecting takes about a minute. You add ContactDrive to your assistant as a custom connector using one URL, sign in to ContactDrive, and choose what the assistant is allowed to do. You don't need an API key.

The URL is the same for every assistant:

https://mcp.contactdrive.io/mcp

Claude (web and desktop)

In Claude, open Settings > Connectors.
Click Add custom connector.
Name it ContactDrive and paste the URL above.
Click Add, then Connect.
A ContactDrive window opens. Sign in if you're asked to, then approve access (see The consent screen below).
In a chat, make sure ContactDrive is switched on in the tools menu, then just ask.

On a Claude Team or Enterprise plan, an organization owner may need to add the connector for the organization first. After that, each person connects with their own ContactDrive sign-in.

ChatGPT

ChatGPT adds custom connectors through developer mode.

In ChatGPT on the web, open Settings > Apps & Connectors.
Under Advanced settings, turn on Developer mode.
Back in Apps & Connectors, click Create.
Name it ContactDrive, paste the URL above as the MCP server URL, and choose OAuth for authentication.
Confirm that you trust the connector, then click Create.
A ContactDrive window opens. Sign in if you're asked to, then approve access (see The consent screen below).
In a new chat, choose developer mode from the tools menu and turn on ContactDrive.

Developer mode is available on ChatGPT's paid plans. In a Business or Enterprise workspace, an admin may need to allow it.

Any other MCP client

Any assistant or tool that supports remote MCP servers with OAuth can connect. Give it the URL above. It will open ContactDrive's sign-in and consent screen in your browser, and once you approve, it's connected. You don't need to register anything or paste a key.

This screen is where you decide what the connection can reach. It's headed Connect to ContactDrive and names the assistant that's asking.

If you aren't already signed in to ContactDrive, you'll sign in first and then come straight back to this screen. It's the usual passwordless sign-in.

Choose workspaces

Under Workspaces, choose an access level for each workspace:

No access. This is the default for every workspace.
Read. The assistant can look things up but change nothing.
Read and write. The assistant can also create and update.

You can set a level per workspace, or use the All workspaces box at the top to cover every workspace you belong to, including ones you join later. Changing any single workspace row switches the all-workspaces choice off.

A workspace whose trial has ended is marked Read-only workspace and can be connected at Read only.

Choose permissions

Under This will allow [assistant] to, you'll see every permission the assistant asked for, grouped by area (Contacts, Activity, Tasks, Email marketing, and so on). Every box starts ticked.

Untick anything you don't want the assistant to have. You can only narrow the list here, never widen it. Some examples of how people narrow it:

A connection for research only: choose Read for the workspace and leave the permissions as they are. Read access only uses the read permissions.
An assistant that should never touch your workspace setup: untick Add and edit custom fields, activity types, tags, and relationship types.

Approve or deny

Click Approve to finish. The button stays disabled until at least one workspace has access and at least one permission is ticked. Click Deny to cancel, and nothing is connected.

Connecting several workspaces

One connection can cover several workspaces. When it does, name the workspace in your request, for example "In the Smith Campaign workspace, find…". If you don't, the assistant will ask. It never guesses, so a change can't land in the wrong workspace. You can also ask "Which ContactDrive workspaces can you see?"

Checking it worked

Ask something simple, such as "What custom fields does my ContactDrive workspace have?" If you get an answer, you're connected. If not, see Troubleshooting Your AI Connection.