Agentic control
Autonomy levels
Use when people need to set how far the agent can go on a task, from watching to acting alone.
Per-task dial for how much the agent may do, observe through autonomous.
Part of Agentic control. The stage where the agent stops answering and starts doing. It covers the whole handover: agreeing what the agent may do, watching it work, stepping in, and checking what changed afterwards. Reach for these as soon as an action touches a customer, money or a record.
Do
- Describe each level by what the agent will do, in one short line.
- List the concrete capabilities and caps of the selected level, such as the daily refund cap.
- Show the current level as a badge wherever the agent acts, so the setting is never hidden.
Don't
- Don't offer a bare on and off switch. Most teams want something between the two.
- Don't let the agent raise its own level. Only a person moves the dial up.
- Don't reuse one setting for every task. Replying and refunding deserve different levels.
States
- Observe: The agent reads and flags, never acts. Badge reads Watching.
- Suggest: The agent drafts and the person sends. Badge reads Suggest only.
- Confirm: The default. The agent acts after an OK, inside a refund cap. Badge reads Asks first.
- Autonomous: The agent acts alone and posts a receipt after each action. The badge pulses and reads Acting solo.
Usage
import { AutonomyPattern } from "./halaska-kit";
<AutonomyPattern />
Built from
Agent Glyph, Badge, Card, Stack, Typography
Related patterns
Halaska UI · Components · Patterns · Screens · Install