AgentWatch

Real-time health & security monitoring for AI coding agents.
Detect loops, stuck states, and security threats before they burn your tokens.

Get Started GitHub

agentwatch check

$ agentwatch check

Agent Health Report  PID 48291 · claude-code · 12m

  Health Score     ████████░░  82/100
  Security Score   ██████████  100/100
  Efficiency       ██████░░░░  64/100

  ⚠ context_pressure  Context 72% full (129K/180K tokens)
  ⚠ high_error_rate   4 consecutive failures detected

  Exit code: 0 (healthy)