autonomous-execution

You can trigger AI workflows by speaking a natural language command such as 'Manu, assess mahg-pms'. The system recognizes your intent and starts the appropriate workflow automatically without needing explicit run commands.

During multi-agent delegation observability, you can see the current run ID, execution phase, active agent, completed tasks versus total tasks, and any blocked tasks. This provides real-time visibility into your workflow execution progress.

Every workflow run creates a persistent record stored in .araya/runs/{run_id}/run.json. This file contains run metadata, start and end times, workflow details, agents, status, artifacts, errors, and the final outcome for long-term audit trails.

Provider-aware routing checks rate-limit history, recent usage, and known failures before delegation. This avoids routing tasks to providers near their quota, ensuring continuous AI workflow orchestration without hitting API rate limits.

To pass the usability gate in autonomous execution, a feature must be usable, solve a real problem, be demonstrable end-to-end, and reduce manual work. These criteria ensure features provide practical value.

No, autonomous-execution does not require explicit run commands to start ARAYA workflows. It applies natural language triggers to recognize intent and initiate execution automatically, making commands optional.