CAUM watches event traces for loops, stagnation, exact cycles, token exposure, and cost exposure while your agents run. It does not read private content, judge correctness, or control the agent.
CAUM is narrow by design. It does not try to be a general tracing suite or a truth detector. It measures whether agent work still has healthy structural movement.
T1-T5 health tiers for each run, with review priority and profile-aware calibration.
Exact cycle coverage, repeated state patterns, deadwork pressure, and loop entry alerts.
Observed token and cost counters are tied to the structural evidence, not to semantic judgment.
Live events can include agent identity and delegation metadata when available. CAUM hashes identity fields in returned evidence and ignores sensitive content fields.
{
"session_id": "deploy-184",
"agent_id": "agent-42",
"identity_sub": "spiffe://acme/prod/agent/42",
"scopes": ["tools:read", "tools:execute"],
"event": {
"event": "tool_call",
"tool": "bash",
"status": "completed",
"input_tokens": 420,
"cost_usd": 0.013
}
}These plans buy structural observability access and onboarding for agent runs. PDF Receipt remains available per upload for after-the-run audit artifacts.
For one builder who wants CAUM on coding-agent runs without an enterprise process.
For a team running agents often enough that repeated loops and cost exposure already matter.
For higher-volume pilots that need a more guided integration path and operator review.
CAUM observes structural evidence only. It does not judge whether an answer is true, does not claim hallucination detection, and does not block agents.
Use Live Meter for running agents, then generate a PDF receipt when a run needs human review, audit evidence, or a customer-safe incident artifact.