Autonomy Settings
Control how much independence your agents have — from full approval to fully autonomous execution.
Autonomy Spectrum
Orbitr gives you control over how independently each agent operates. You can configure this per-agent and per-task-type to match your comfort level.
Modes
Supervised (Default)
Every task requires your explicit approval before execution. The agent researches, plans, and prepares — but waits for your go-ahead.
Best for: New users getting comfortable with agents, high-risk actions, brand-sensitive content.
Semi-Autonomous
Low-risk tasks execute automatically. Higher-risk tasks still require approval. You define what counts as "low risk" based on task type and potential impact.
Best for: Experienced users who trust their agents for routine work but want oversight on significant changes.
Fully Autonomous
All tasks execute without waiting for approval. You can still review and revert after the fact, but agents don't wait.
Best for: Speed-focused users who want maximum throughput and trust their agent configuration.
Configuring Autonomy
- Navigate to Settings for your site
- Find the autonomy controls for each product
- Choose your preferred mode
- Optionally, set per-task-type overrides
Task-Type Granularity
You can mix autonomy levels within a single agent. For example, with the SEO Agent:
| Task Type | Autonomy Level |
|---|---|
| Fix broken links | Autonomous |
| Update meta descriptions | Autonomous |
| Create new pages | Requires approval |
| Change URL structure | Requires approval |
This gives you the best of both worlds — speed for safe operations and oversight for impactful changes.
Safety Rails
Regardless of autonomy settings, certain safety mechanisms are always active:
- Budget caps — Agents never exceed configured spending limits
- Rate limiting — Agents won't flood your site with changes
- Rollback capability — Changes can be reverted if something goes wrong
- Audit trail — Every action is logged, whether autonomous or approved
Changing Settings
You can change autonomy settings at any time. Switching from autonomous to supervised immediately starts queuing new tasks for approval. Switching the other direction immediately processes any queued tasks that were waiting.
Recommendation
Start in supervised mode for the first few weeks. This lets you see what each agent does and build confidence in its decision-making. Then gradually enable autonomy for the task types where you're consistently approving without changes.