Agent Execution

Learn how agents execute tasks and how to get the best results from AI workers.

Running Agents

There are two ways to run agents on a node:

From the Map or List View

1. Select a node (project, task, etc.)
2. Click the Run button (play icon) on the node
3. Choose an agent from the dropdown menu
4. The agent will begin execution

From the Edit Dialog

1. Open a node's edit dialog by clicking on it
2. Navigate to the Agents section
3. Select an agent and click Run
4. View execution progress and results directly in the dialog

The edit dialog is also where you can view results from previous agent executions, making it easy to review what agents have produced.

The Execution Lifecycle

Agent Selected → Agent Starts → Working → Output Generated → Review → Complete

1. Agent Selection

When you run an agent on a node:

• The agent receives node context (name, description, files)
• Agent-specific settings are applied
• Execution is queued

2. Execution Begins

The agent starts working:

• Status changes to "In Progress"
• Real-time updates appear in the interface
• You can monitor progress

3. Output Generation

The agent produces results:

• Generated content appears in the output panel
• Intermediate results may be shown
• Files may be created or modified

4. Review Phase

Before finalizing:

• Review all generated content
• Make edits if needed
• Provide feedback

5. Completion

Once approved:

• Task status updates to "Complete"
• Output is saved
• Next steps may be suggested

Monitoring Execution

Status Indicators

Status	Meaning
⏳ Queued	Waiting to start
🔄 In Progress	Agent is working
⏸️ Paused	Temporarily stopped
✅ Complete	Successfully finished
❌ Failed	Error occurred

Progress Panel

The progress panel shows:

• Current step being executed
• Time elapsed
• Resource usage
• Intermediate outputs

Handling Issues

If an Agent Fails

1. Check the error message
2. Review task description for clarity
3. Verify required files are attached
4. Try again or adjust settings

If Output Is Unsatisfactory

1. Review the task description
2. Add more context or examples
3. Adjust agent settings
4. Request regeneration

Best Practices

Before Execution

• Provide clear, detailed task descriptions
• Attach all relevant files
• Set appropriate agent parameters
• Define success criteria

During Execution

• Monitor progress periodically
• Check intermediate outputs
• Be ready to provide clarification

After Execution

• Thoroughly review all output
• Test generated code/content
• Document any issues
• Provide feedback for improvement

Cost Considerations

Agent execution may incur costs:

• AI model usage (tokens processed)
• External API calls
• Storage for generated files

Monitor usage in the Cost & Value layer to track spending.