← All agents
Platform & control

Agent Architect

LLM agentadvanced (Opus)Added, untested

Meta-agent that diagnoses and proposes fixes to other agent definitions during a weekly health review.

Book a demo →Review controls
WHAT IT DOES
  • Reviews failing agents and diagnoses the definition or configuration cause.
  • Proposes fixes to agent definitions.
  • Runs a weekly agent-health review.
TRIGGER EVENTS

The agent does not run on its own. It wakes only when one of these events lands in the governed event queue.

agent_instance.failedschedule.triggered.agent_health_review
CONTROL

Failure events skip the architect itself, so it can never enter an infinite failure loop. Uses the advanced model tier for the reasoning involved.

See Agent Architect run on your books.

Every Artifi agent operates the same way: it listens for an event, works through governed tools, and writes only what a human approves — all inside Claude.

Book a demo →