Skip to main content
The Task Kanban tracks all agentic task executions across your organization. This page covers common patterns and workflows for using the Kanban effectively.

​
Monitoring Agent Executions

​
Real-Time Visibility

When you execute an agent via Meta Agent, API, or SDK, the Kanban provides real-time visibility: 
"Execute the Security Engineer agent to scan for CVEs in production"
What you see on the Kanban:
  • Task card appears in Pending column
  • Moves to Running when a worker picks it up
  • Live streaming of agent reasoning and tool calls
  • Final state: Completed or Failed

​
Multi-Agent Workflows

Track complex workflows spanning multiple agents: 
"Run a security audit: scan for CVEs, analyze findings, create remediation PRs"
Kanban shows:
  • Parent task with child executions
  • Each agent’s progress independently tracked
  • Dependencies between tasks visualized
  • Overall workflow status

​
Debugging Failed Executions

​
Investigating Failures

When a task fails, the Kanban provides debugging tools:
  1. Click the failed task to open details
  2. Review tool calls to see which step failed
  3. Check logs for error messages
  4. Examine context (environment, queue, agent config)
Common failure reasons:
  • Worker connectivity issues
  • Policy violations blocking actions
  • Missing credentials or permissions
  • Agent tool execution errors

​
Retrying Tasks

From the task details panel:
  • Retry - Re-execute with same parameters
  • Retry with changes - Modify task before re-execution
  • Clone - Create a new task based on this one

​
Tracking Executions by Source

​
Meta Agent Executions

Tasks initiated via Meta Agent chat:
  • Show the original natural language prompt
  • Include conversation context
  • Link back to the Meta Agent session

​
API Executions

Tasks initiated via REST API:
  • Show API request details
  • Include caller information
  • Useful for automation pipelines

​
SDK Executions

Tasks initiated via Python/Go SDK:
  • Show SDK client details
  • Include application context
  • Common in CI/CD integrations

​
Filter by Status

Focus on specific task states:
  • Running - Active executions needing monitoring
  • Failed - Tasks requiring investigation
  • Completed - Successful executions for audit

​
Filter by Agent

See all executions for a specific agent:
  • DevOps Engineer tasks
  • Security Engineer tasks
  • Custom agent tasks

​
Filter by Queue

Track executions on specific infrastructure:
  • Production queue tasks
  • Dev queue tasks
  • GPU worker tasks

​
Filter by Time

Review historical executions:
  • Last hour
  • Last 24 hours
  • Custom date range

​
Integration Patterns

​
CI/CD Pipeline Tracking

Track agent executions triggered by CI/CD: 
# GitHub Actions example
- name: Execute deployment agent
  run: |
    kubiya exec "Deploy to staging" --queue staging-workers
All pipeline-triggered tasks visible on Kanban.

​
Scheduled Job Monitoring

Background Jobs execute agents on schedules:
  • Each scheduled run appears on Kanban
  • Track success/failure rates over time
  • Debug recurring issues

​
Webhook-Triggered Tasks

External events can trigger agent executions:
  • Alert webhooks → incident response agents
  • PR webhooks → code review agents
  • Deployment webhooks → verification agents
All webhook-triggered tasks tracked on Kanban.

​
Audit and Compliance

​
Execution History

The Kanban maintains full execution history:
  • Who initiated the task
  • What agent executed
  • Which worker processed it
  • All tool calls and outputs
  • Final results and artifacts

​
Export and Reporting

Export execution data for compliance:
  • CSV export of task history
  • API access to execution logs
  • Integration with audit systems

​
Best Practices

  1. Monitor running tasks - Check the Kanban for stuck or long-running executions
  2. Investigate failures promptly - Failed tasks may indicate infrastructure issues
  3. Use filters effectively - Focus on relevant tasks for your role
  4. Review execution patterns - Identify recurring issues or optimization opportunities
  5. Set up alerts - Configure notifications for failed executions