Choose an Autonomy Mode
ClawPaw separates intelligence from authority. Just because it understands what should happen does not mean it should immediately act. Autonomy modes exist to give you explicit control over how decisions translate into system changes.
Each task runs under a selected execution mode. You can change modes at any time, and different tasks can operate under different levels of autonomy.
The Three Modes
ClawPaw supports three execution modes:
Explore
Explore Mode is read-only. ClawPaw analyzes your environment and generates plans, summaries, or structured insights without modifying files or system state.
In Explore Mode, ClawPaw can:
- Inspect folder structures
- Read approved documents
- Analyze browser content
- Propose step-by-step plans
It cannot:
- Modify files
- Send messages
- Update calendar events
- Perform system actions
This mode is ideal for planning, auditing, and reviewing before committing to changes.
Ask to Edit
Ask to Edit Mode introduces collaborative execution. ClawPaw prepares intended modifications and presents them clearly before applying them.
In this mode, ClawPaw will:
- Show the exact changes it plans to make
- Request confirmation before executing
- Allow adjustments before approval
This is well-suited for file restructuring, content rewriting, scheduling changes, or multi-step workflows where review matters.
Execute
Execute Mode allows ClawPaw to perform approved actions directly within your granted permissions and folder scope.
Even in Execute Mode, ClawPaw:
- Cannot access unapproved directories
- Cannot bypass macOS privacy controls
- Cannot escalate privileges
Execution remains bounded by system permissions and sandbox constraints.
Why Modes Matter
Many autonomous agents treat intelligence and execution as inseparable. ClawPaw intentionally decouples them.
This design provides:
- Reduced risk
- Clear authority boundaries
- Adjustable control per task
Autonomy is not global. It is contextual.
ClawPaw understands first. It acts only within the limits you define.