Skip to main content
Agent Mode is Verdent’s default execution mode where the agent takes direct action on your instructions. Rather than structured planning sessions, Agent Mode prioritizes quick execution while maintaining full transparency.

What You’ll Learn

  • When Agent Mode is the right choice
  • How the execution workflow operates
  • Managing permissions and visibility

When Agent Mode is Useful

Agent Mode works best when you have clarity on what needs to be done:
ScenarioSituationHow Verdent Helps
Clear RequirementsYou know exactly what you want implemented. The task is well-defined with a specific outcome.Agent proceeds directly with implementation. No planning overhead, just results.
Quick IterationsYou’re in a rapid feedback loop, making incremental changes based on results.Fast execution cycles let you see changes immediately and adjust course.
Straightforward TasksBug fixes, small features, code cleanup, or refactoring with obvious scope.Agent completes the work efficiently without unnecessary back-and-forth.

Agent Mode Workflow

1

Submit Your Request

Describe what you want done. Clear, specific instructions work best in Agent Mode.
2

Agent Takes Action

Verdent begins implementation immediately. File edits, tool usage, and progress appear in real-time.
3

Permission Prompts

Certain actions require approval. Review and approve to proceed.
4

See Results

Work completes with full visibility into what changed and why.

Key Characteristics

CharacteristicDescription
Default ModeNew tasks start in Agent Mode unless you switch to Plan Mode
Direct ActionAgent proceeds with implementation rather than extensive planning
Full VisibilityFile edits, tool usage, and progress are shown in real-time
Inline ClarificationQuestions appear naturally in the chat, not as structured questionnaires

Agent Mode vs Plan Mode

Choose based on task clarity and complexity:
Agent Mode is optimal for:
  • Clear, well-defined requirements
  • Straightforward implementation tasks
  • Quick iterations and feedback loops
  • Tasks where you’re confident in the approach
  • Bug fixes and small features

Permissions in Agent Mode

Agent Mode executes tasks directly, but certain actions still require explicit approval:
  • File modifications outside expected scope
  • Destructive operations
  • Certain tool executions
  • Actions with significant side effects
Permission prompts protect against unintended changes while maintaining execution speed for routine operations.

FAQs

Yes. New tasks start in Agent Mode. Switch to Plan Mode when you need structured planning.
Yes, when needed. Questions appear inline in the chat rather than as structured questionnaires. The agent may proceed without asking if the task is sufficiently clear.
Press Shift+Tab or Cmd+. to toggle modes. Mode switching is available at certain points, not during active execution.
No. Certain actions still require approval. Agent Mode reduces friction for routine operations while maintaining safety for significant changes.
Use Plan Mode for complex features, architectural decisions, or when you want to explore and refine requirements before implementation begins.

See Also