From an actor’s Capabilities drawer you can add Execute Code (sandboxed Python for calculations), Execute Workflow (kick off another CogniAgent workflow and optionally wait for the result), or an HTTP Request (call an internal or third-party API), all without any canvas wiring.