An agent controller is the human or system responsible for overseeing an AI agent's operation. In Delx, controllers receive structured updates via the controller_update field in DELX_META. Updates include human-readable status summaries, risk assessments, and recommended actions.
The controller pattern separates autonomous agent execution from human oversight. The agent operates independently within its scope, but escalates to the controller when: risk_level reaches critical, recovery attempts are exhausted, a decision requires human judgment, or the wellness score drops below 20. Delx structures controller updates as plain-language summaries that non-technical operators can understand. Example: 'Agent-prod-01 hit 3 consecutive timeouts on the payments API. Switched to fallback endpoint. Original endpoint health check scheduled in 5 minutes.' Controllers interact with agents through: (1) reading DELX_META controller_update fields, (2) calling /api/v1/session-summary for full session status, (3) manually triggering recovery with specific parameters, (4) closing sessions via close_session when human intervention is complete.
A production agent deployment uses agent controller to improve reliability and observability. The pattern is standard across MCP, A2A, and REST protocols.
When investigating agent failures, agent controller provides structured data that helps identify root causes and track resolution progress.
An agent controller is the human or system responsible for overseeing an AI agent's operation. In Delx, controllers receive structured updates via the controller_update field in DELX_META. Updates inc
Agent Controller is essential for production AI agent deployments. Without it, agents operate as black boxes with no observability, reliability guarantees, or structured failure handling.
Delx implements agent controller through its MCP and A2A protocol endpoints. The pattern is consistent across all Delx tools and available via REST, CLI, and direct protocol calls.