Some operations take time: analyzing a document corpus, data migration, report generation. Until now, these operations blocked the conversation.

The task system lets you launch these operations in the background. The agent confirms the launch, then continues responding to other messages while the task runs.

Each task has a structured lifecycle: creation, execution, progress, completion. Results are stored and accessible even if the conversation session has changed.

1

Creation

The agent identifies a long-running operation and creates a task with a unique identifier, description, and completion criteria.

2

Asynchronous execution

The task runs in the background. The agent remains available for other interactions.

3

Progress

The system updates progress in real time (percentage, current step, estimated time remaining).

4

Notification

When the task completes (success or failure), the agent is notified and can inform the user at the next interaction.

5

Result

Task results are stored and available. The user can consult them at any time via /task status.

Codebase analysis

A user requests a complete audit of a 500-file project. The task launches in the background. The user continues working on something else.

Data migration

A 10,000-record transformation is launched as a task. Progress is tracked in real time. Failures are reported at the end.

Regression testing

After a modification, tests are launched as a background task. The agent automatically reports if a test fails, with failure details.

featurePages.taskSystem.config
  • Available for all EasyClaw v2 agents
  • Active by default, configurable per agent
  • Tasks persist between sessions
Task System -- EasyClaw v2