Skip to main content
The Tasks section lets you inspect every individual task executed within a job. Tasks represent specific interactions handled by agents, including AI conversations, media display, tool invocations and more.

​
Tasks List Overview

The top section displays a searchable and filterable list of all tasks, with key information like:
  • Task Key: Unique ID for the task.
  • Action: Linked action (e.g., “Talk to AI Agent”, “MCP Agent”).
  • Job Key: Parent job the task belongs to.
  • Created On: Timestamp of task creation.
  • Status: Task progress (e.g., STARTED).
Task List Each row can be expanded to reveal full task details.

​
Task Details Panel

Upon expanding a task, you can review:
  • Input: Parameters passed into the task (if any).
  • Output: Execution results including, as example:
    • Transcripts
    • Token usage
    • TTS/STT durations
    • Model and provider details
  • Config: System-level runtime configuration of the task (useful for debugging).
  • Timeline: Creation and start timestamps.
  • Task Actions: A real-time log of tool executions (e.g., show_media_to_user) with origin/destination info and timestamps.
Task Details

​
✅ Common Use Cases

  • 🧪 Debugging tool behavior (see which tool was called, when and by whom).
  • 📊 Reviewing AI output (e.g., conversation transcripts and LLM statistics).
  • 🛠 Understanding configuration passed to the agent during execution.
  • 👀 Auditing real-time activity and ensuring workflow correctness.
Tasks are the atomic building blocks of the job execution system. They give full observability into what happened, when and how.