Connect an agent
Open Agents in the intended workspace and choose Add agent. A workspace owner or administrator manages agent connections.
Choose a connection
Section titled “Choose a connection”- Claude Code: follow the installation commands, run
/actera:connect, and approve the connection in the browser. Check that the displayed codes match. See the Claude Code guide. - API or SDK: name the agent, choose TypeScript, Go, or HTTP, and select Create key. Store the one-time API key in your application’s secret manager. Use the API address and example shown in setup.
The connection belongs to one agent in one workspace. HTTP decision requests can omit the agent identity because the key supplies it. SDK clients still bind to the exact agent ID shown in the generated example.
Send a first decision request. The dialog updates when it arrives. If you close setup, you can return to the agent under Not set up.
Set its authority
Section titled “Set its authority”Choose published policies based on the actions the agent sends. Review When no policy matches, then save the configuration.
New agents default to Allow for unmatched requests. Select Block if every permitted action must be covered by an explicit rule. Assigning a policy does not guarantee that every request matches one of its rules.
Claude Code starts in Monitoring. For an SDK or HTTP integration, set the mode explicitly during rollout. Follow Start from observed traffic before switching to enforcement.
Manage the connection
Section titled “Manage the connection”Open the agent’s page and use Connection to see whether it is connected, waiting for a request, or needs a key.
Each agent has one active connection. API keys expire after 90 days. Replace key invalidates the old key immediately; update the application secret as part of the change. Keys are shown once.
Disconnect immediately stops the current key from authenticating. The agent and its policies remain. To reconnect Claude Code, run /actera:connect again. For an API connection, create a new key from the agent’s page.
A successful readiness check confirms that the connection can send requests. It does not verify policy coverage or that your application enforces decisions.
Check Plan & usage if you cannot add another agent. Registered agents count toward the workspace allowance even before setup is complete.