Agents
Route: /agents · Detail: /agents/$id
What it is
An AI worker with a mission, model, skills, and permissions. Agents do the actual work — drafts, updates, tool calls — under the rules of their department.
What it is for
- Hire or create workers for recurring outcomes
- Start / pause / stop lifecycle
- Watch a live run, chat, configure tools and skills
- Narrow permissions below the department frame
Where you are in the flow
Department exists + model + capas → ★ Agent configured → Run / schedule / chat
→ maybe My work → done (Activity)
Tabs on the detail page
| Tab | Purpose |
|---|---|
| Overview | Status, mission, current task |
| Live Log | Streaming transcript of the active run |
| Chat | Talk to the agent / send a task |
| Files | Workspace files for this agent |
| Configuration | Mission, model, triggers, autonomy |
| Access & Permissions | Tool narrowing vs department defaults |
| Skills | Which procedures this agent may load |
| Memory | What it is allowed to remember |
| History | Past runs |
What you do here
- Create an agent or hire from a capa template (Capas).
- Assign a model, mission, and skills.
- Start the agent; send a low-risk task.
- If it waits — open My work.
- Review Activity / History after.
Where work goes next
| From agent | Next |
|---|---|
| Needs approval | My work |
| Hands to another team | Handoffs |
| Part of a pipeline | Flows |
| Missing tools | Capas → department permissions |
Isolation note
When isolation is on, each run can execute in a sandbox container. Details: Key concepts — Agent architecture.