Agent task management

Agent task management sequences in workflow processes define how agents dynamically plan, run, and track multi-stage projects when task planning is enabled.

Agents are programmed to use the built-in tools proactively when navigating complex, multi-stage projects in the following sequence:

Task creation — Before starting a run, the Agent plans the workflow by creating distinct tasks for each required step.
Task in progress — The Agent updates a task status to in_progress before running that specific step.
Task completion — The Agent marks the task as completed upon finishing the step.

For single-step or simple requests, the Agent skips this task planning sequence entirely and provides a direct response.

Agents are instructed not to distinguish between the completion of a to-do and the overall request. While task completion is a necessary log for planning and coordination, the final delivery of the solution remains a separate and concluding phase of the workflow.

The task list is dynamic and not predefined. Agents generate these lists during runtime, based on the specific details of a request. For example, a request to write a report on Q1 sales results in a different set of tasks than a request to update the database schema.

UI visibility and tracking

During a workflow process, a dedicated To-Dos card within each agent panel is displayed to monitor progress using the following tracking elements:

Progress tracking — A completion counter displays the ratio of finished tasks relative to the total number of generated tasks. for example, 2/5 done.
Task status indicators — The card lists each generated step with a visual icon representing its current operational state:
- ⚪ Pending
- 🔄 In progress (highlighted in blue)
- ✅ Completed (shown with strikethrough)
Figure 1. Real-time task tracking in Agent Studio