Running agents

Start the supervisor, run an installed agent, and watch its status, health, and logs.

The runtime supervisor runs locally. Start it once, then run any installed agent under it.

$ norien runtime start$ norien run research-agent  ✓ running · pid 24180 · healthy$ norien status               # status + health of everything$ norien logs research-agent -f$ norien restart research-agent$ norien stop research-agent

Reading the state

status shows two axes per agent. Status is what the supervisor is doing — running, stopped, failed, restarting. Health is what the agent reports — healthy, unhealthy. An agent can be running and unhealthy; that is a real state worth seeing.

Recovery

A crash is restarted automatically, with a loop cap: an agent that keeps failing is left stopped rather than spun forever. Environment variables the agent needs are injected at launch — including secrets, which are never written to logs.

CarefulBecause it executes code, the supervisor is local-only. The hosted app's runtime page shows the live registry and chain status and points you here to start it yourself.