Supervision (agent-body)

The entire Autonomic ecosystem runs on a foundation of reliability and daemonization provided by agent-body. Rather than manually starting individual servers in separate terminal tabs, agent-body acts as the master supervisor process.

The Process Tree

When you run autonomic start, agent-body launches as the root of the process tree. It reads the local environment and begins daemonizing the 8 other organs according to a strict boot sequence:

NATS Server (agent-nerves foundation): Booted first to establish the event bus.
agent-brain: Booted second to provide memory and context routing.
agent-spine, agent-heart, agent-muscle, etc.: Booted concurrently once the bus and brain are healthy.

Health Checks & Self-Healing

The core concept of agent-body is resilience through isolation. Because each organ is a completely isolated binary, a failure in one does not cascade.

agent-body continuously monitors the PID and /health HTTP endpoint of every organ it supervises. If agent-eyes runs out of memory while processing a massive screenshot, the OS will kill it. agent-body immediately detects the termination, logs the crash, and spawns a fresh agent-eyes instance. The new instance simply reconnects to the agent-nerves event bus and picks up the next task from the JetStream queue.

Standardized Logging

agent-body intercepts the stdout and stderr streams of all child organs. It formats them into a unified, color-coded stream that you can tail using autonomic logs. This provides a centralized view of the entire distributed system without needing to hunt for individual log files.