A practical guide

Give the Agent the right connection.

Choose an account and a useful scope.

All guides
1

Start with the task

Decide what the Agent should do in the app: read a document, prepare a draft, inspect an issue, or update an approved resource. Browse Integrations for the supported connection type.

2

Choose whose account is involved

A connected Google account, a customer-installed Slack or Teams bot, and an Agent-owned browser account represent different identities. Use the one intended for the work.

3

Complete the supported setup

Follow the app connection flow or ask the Agent to help you open it. When a browser sign-in is needed, complete it on the Agent’s desktop. Enter credentials into the service’s sign-in page rather than the conversation.

4

Review access

Grant the accounts, repositories, stores, or conversations needed for the job. A connection makes an integration available; the Agent still needs the relevant permissions.

5

Try a small first task

Ask for a read-only check, such as finding a named document or summarizing an issue you chose. Review the result before expanding the task or enabling actions.

Make a start

Start with one agent and a job worth doing.

Create a workspace, bring your team, and see what you can get done.