Skip to main content
A Computer Agent needs both published Agent instructions and a provisioned Computer before it can perform work.

​
1. Create or select an Agent

Choose the Agent account that will own the Computer. If the Computer already hosts other Agents, you can add another Agent from the organization’s Computer settings. Give each Agent a clear name and responsibility so Tasks, channels, browser access, and results are assigned to the right account.

​
2. Configure the Agent

Open Edit Agent and define:
  • the Agent’s role and operating boundaries
  • the outcomes it is responsible for
  • when it must ask for approval
  • how it should report completed, blocked, and failed work
  • the intelligence model appropriate for the work
Test and publish the configuration before relying on it for ongoing operations.

​
3. Provision the Computer

Open the Computer settings and provision the Agent. Provisioning can take a few minutes. Wait until the Computer reports that it is ready before connecting capabilities. If a usable Computer already exists for the organization, add the Agent to that Computer instead of creating an unnecessary additional deployment.

​
4. Add only the capabilities it needs

Install skills, connect channels, sign in to required websites, add managed credentials, and upload reference files for the initial workflow. Avoid granting unrelated access. See Computer Agent capabilities for the available building blocks.

​
5. Run a safe test

Assign one read-only or reversible task. Review the linked work conversation, including the Agent’s tool activity and final result. Independently verify the result in the destination system. Test failure and approval behavior before enabling repeating work, heartbeats, or customer-facing delegation.