Skip to main content
Connectors give Primo AI access to tools outside the platform. A connector is a Model Context Protocol (MCP) server that you register by URL. Once connected, the assistant can call that server’s tools during a conversation, for example to look up a page in Notion or act in another system.

​
When to use a connector

Add a connector when you want the assistant to reach beyond Primo’s own data:
  • Query another system — pull information from a tool your team already uses.
  • Act in another system — let the assistant perform tasks in a connected service.
  • Extend the assistant — bring in any tool that speaks the Model Context Protocol.
If you only need the assistant to work with Primo data or your company’s own instructions, you don’t need a connector: use knowledge base items instead.

​
Prerequisites

  • The Manage AI agents permission to add, edit, re-authenticate, or delete connectors, and to change their tool permissions. View AI agents is enough to see them.
  • The URL of an MCP server that supports OAuth authentication.

​
Add a connector

1

Go to AI Agents > Settings

Open AI Agents > Settings, then the Connectors tab.
2

Click 'Add connector'

The connector dialog opens.
3

Enter the connector details

  • Name — a label for the connector, for example Notion.
  • Server URL — the MCP endpoint, for example https://example.com/mcp.
4

Add advanced settings if needed

Expand Advanced settings only if your MCP server does not support Dynamic Client Registration. Register Primo in the provider’s OAuth console, then enter the Client ID and Client secret. Leave the secret empty for public clients that use PKCE.
5

Click 'Connect'

Primo redirects you to authenticate with the MCP server. After you authorize access, the connector returns to Primo and its status becomes Connected.

​
Connector statuses

​
View a connector

The Connectors tab lists the built-in connectors, Primo and Primo Help Center, then the connectors you added. Built-in connectors are always Connected. Click a connector to open its page. For a connector you added, the page shows:
  • Settings — the Name, the Server URL, and the Client ID if you set one. Primo never shows the client secret. Click Edit connection to change them. A new server URL or new OAuth credentials mean authenticating again.
  • Chat permissions and Agent permissions — the tools the MCP server exposes, with the policy of each. See Set tool permissions.
Primo reads the tools from the MCP server when you open the page, so the list follows what the server exposes. Tools appear only while the connector is Connected. If the server doesn’t answer, a warning replaces them and Settings stays available. A built-in connector’s page has no Settings and no Actions menu. You can’t edit or delete it, but you can set its tool permissions.

​
Manage a connector

On the page of a connector you added, open the Actions menu at the top right:
  • Re-authenticate — sign in to the MCP server again. It appears only when the connector isn’t Connected.
  • Delete — remove the connector and its stored credentials, after you confirm. You can’t delete a connector while an agent uses it: turn it off in that agent first.

​
Set tool permissions

A connector’s page sets how Primo AI uses its tools. Once the connector is Connected, its tools appear in two sections:
  • Chat permissions apply to your conversations with Primo AI.
  • Agent permissions apply when Primo AI works on its own, for example on a ticket you add it to as agent. A custom agent copies them when you turn the connector on for it, then keeps its own copy: later changes here don’t reach it.
Each section splits the tools into Read-only tools and Read-write tools. A tool is read-only only when the MCP server marks it with the readOnlyHint annotation. Every other tool counts as read-write. Set a policy for a whole group from the menu on the right of the group. It also covers tools the server adds later, and it replaces the policies you set tool by tool in that group. To set a single tool, expand the group and pick its policy with the icons next to its name. A group whose tools have different policies shows Custom. Until you change them, the policies are: Agents can use read-write tools from the start. To keep agents read-only on a connector, set Read-write tools to Always refuse under Agent permissions. A custom agent that already uses the connector keeps its own copy, so change it in the agent too.

​
How Primo AI uses connectors

In a conversation, Primo AI can use the tools of every connected connector and picks the one a request needs. To point it at a connector, type @ in the message box and select the connector. Before it runs a tool, Primo AI follows the connector’s Chat permissions: by default, it asks you before it uses a tool that writes or changes data. See Primo AI.