chat_with_agent

To delegate a specialized review, call list_agents() to discover an available agent, extract its ID, then use chat_with_agent() or submit_to_agent() to send the request. This approach enables obtaining specialized judgment or a second opinion without losing your existing context.

Agent orchestration handles multi-turn session continuity by passing an optional session_id parameter to chat_with_agent(). This preserves multi-turn context, allowing you to continue conversations or resume prior sessions seamlessly across different agent interactions.

Foreground agent task submission uses chat_with_agent() for immediate replies and live help, while background submission uses submit_to_agent() for longer tasks. You then monitor background job progress by calling check_agent_task() to retrieve status updates.

Yes, you can submit a long-running background job to another agent using submit_to_agent(), which allows the work to be processed asynchronously. You can then monitor the progress by calling check_agent_task() to retrieve the current task status.

You should use agent delegation when tasks require specialized judgment, a second opinion, or expert review of a technical design. It is also applicable when a user explicitly requests a specific agent or when work must be submitted and monitored as a background job.