Agents
The Agents screens - definitions, instances, runs and the run log for background AI workers.
Agents run in the background, read your data with the access of the person who created them, and email people on your team or write events and attributes. The Agents screen in the Analyze group of the navigation rail is where you define them, run them and check what they did. If you are new to the idea, read Optima Agents first.
The screen needs the Agents licence package (which requires AI assistants), with the features Custom agents and Built-in agents. The Agents entry is missing from the navigation when the licence has neither feature. The Definitions tab and New definition need Custom agents; the Built-in tab needs Built-in agents. Reading needs agents_editor, agents_auditor or auditor, or an administrator role; see licences and roles. Agents need a text model from AI models.
| Screen | Use it to |
|---|---|
| Built-in agents | Pick a ready-made agent maintained by Optima and create an instance of it. |
| Definitions | Write what an agent does: prompt, triggers, tools, actions and parameters. Each save is a new version. |
| Instances | See every running agent, and create an instance of a definition with a model and a budget. |
| Instance detail | Start, stop or switch one instance between dry run and live, and watch its budget. |
| Run log | Read the runs of an instance and the log of everything the agent did and why. |
| Continuous agents | Set up, preview and start an agent that sweeps a segment or an event type. |