Your agents
An agent is the thing that does the work. It runs on a computer you've connected — your laptop, a Raspberry Pi, or a cloud computer we run for you.
Where it lives
Sidebar → Agents.
Adding one
+ Add agent puts a new agent on a device you've already connected — no re-pairing, and your existing agents stay put. Give it a name and say what it's for.
If you have no device yet, Pair a new device instead takes you there.
Opening one
Click an agent for its tabs:
- Overview — how many tasks it's run, its success rate, which device it's on,
Set as default agent, which projects it works for, and Test connection.
- Model — which model it plans with. High-stakes tasks are upgraded
automatically regardless.
- Subagents — the specialists it can hand work to.
- Tools — which connectors it can reach.
- Access — who else may use it.
Binding an agent to projects
On Overview, Assign to projects limits an agent to one project and everything under it. That's a real boundary, not a hint: a task outside that subtree is refused rather than quietly run.