Core
Connectors
Connect your tools so the agent can read data and collect evidence, and control what it may change.
Connectors link the agent to your other tools: work apps, cloud services, developer and security tools. Once a tool is connected, the agent can use it in a session, in a routine or to collect evidence. "Connect your tools so the agent can collect compliance evidence automatically."
Screenshot
The directory
Open the Connectors page. The page lists what is connected and a directory of available tools. Search with "Search connectors...", filter by All, Work apps, Cloud, Dev & data or Security & identity, and open Browse directory to see the full catalogue. The catalogue groups tools in categories such as Communication, Documents & Knowledge, Project Management, CRM & Sales, Finance & Payments, HR & People, Cloud, Database, Security & Identity and Compliance. "Show recommended set" returns to the short list.
In short
Try it
Open Connectors and pick a tool from the directory, for example GitHub or Slack.
Choose Connect. For most tools a sign-in window opens ("You'll be redirected to sign in and authorize access, no token to copy"). Finish signing in there; the panel updates on its own.
The tool now appears under Connected. Open it to see its status, its tools and View Logs.
Good to know
- If your browser blocks the sign-in popup, use the button shown to open the window.
- "This connector isn't configured on this instance yet": an administrator has to register the app before it can be connected.
- Credentials are encrypted at rest and isolated per tenant.
Add a custom connector
Add connector offers Browse directory or Custom. A custom connector is a MCP Server (enter the "Remote MCP server URL") or a REST API (enter the Base URL). Give it a name and a description, then choose the Authentication: None, API Key, Bearer Token or OAuth 2.0 (with optional client ID and secret). Test connection checks it, and Sync Tools reads the tools the server offers.
Only add servers you trust
Good to know
- You need the right to manage connectors. The permissions are View (status, tools and logs), Execute (run tools and trigger actions) and Manage (install, configure and remove).
- Remove a connector with Remove; the app confirms with "Connector removed".
Enable, disable and disconnect
A connected connector can be Enabled or Disabled without losing its connection ("Connector enabled", "Connector disabled"), or Disconnected. A connector that stopped working shows "Failed" with the last check time.
Good to know
- Disabling keeps the connection; disconnecting removes it.
Allow write
By default the agent may only read. Allow write "lets the agent create or modify data in this connector. Off by default." Switching it on may ask you to sign in again to grant write access ("Write access enabled"). Individual tools can be set to Always allow or Needs approval.
In short
Connectors in a session
In the composer you can add a connector as context or mention it with @. Each session can use a chosen set of connectors: "name is part of this session, click to remove" and "name is not part of this session, click to add". The messages show which connector was used ("Used name"), or that it failed. See The AI agent.
GitHub repository access
For GitHub you choose the repositories: "The assistant reads only the repositories switched on here." Manage repositories on GitHub opens GitHub in a new tab to add or remove repositories, then brings you back. Two levers apply: what GitHub grants, and what you switch on in the app.
Installed connectors and settings
The route Connectors, Installed lists the connectors your organization added ("None added in this org" if there are none) with type, status, last check and tools. The same management is available in Settings under Connectors.
Good to know
- Reports from the agent can post to Slack under your own name when you choose Connect my Slack; Disconnect mine removes it.
- Whether routines can call a connector is described in Routines.