Events and configuration

​

Event families

• Agent start, end, pause, and resume
• Model request and response
• Tool use and result
• Hook triggered and completed
• Human wait, input, pause, and interrupt
• Explicit errors

​

Inspect an event contract

• Dashboard
• CLI

1. Open Observe → Events.
2. Set a short time window and filter by environment, agent, and event type.
3. Select an event to inspect its normalized fields and raw payload, then open its session to see the execution context.
4. If a paired duration is missing, verify that the start and completion events use the same correlation ID.

Use the Events stream to narrow the data to one agent run and inspect the normalized event fields.Then follow the event into its session. The trace shows what happened immediately before and after it, which is necessary when the payload alone is ambiguous.Compare correlation IDs and timestamps across both views when a paired event or duration is missing.

Discover valid filter values, then fetch full event payloads for one session:

fp list event_types
fp list envs
fp events --since 1h \
  --env production \
  --event-type tool_use,tool_result \
  --session-id <session-id> \
  --full

Use fp --json events ... --fields ts,event_type,session_id,payload for machine-readable inspection.

​

Reserved SDK fields

​

Local boundaries

• Failproof AI state lives under ~/.failproofai unless explicitly configured otherwise.
• failproofai-sdk uses AGENTEYE_HOME for its legacy base-directory override.
• Environment labels can be set through SDK configuration or AGENTEYE_ENVIRONMENT.
• Daemon credentials are stored separately from non-secret settings.

​

Change an environment label

export AGENTEYE_ENVIRONMENT=production-us-east
failproofai config
failproofai config --status
fp sessions --since 1h --env production-us-east