Start here if you are new to CAUM. Inspect real reports and vaults, then run your own trace or start CAUM Live. CAUM observes structure only: no semantic truth scoring, no content judgment, no agent blocking.
CAUM now has two commercial entry points: one receipt for one trace, or Live for recurring agent workflows. Evidence Kit, Evidence Layer, dashboards, and APIs are parts of those two products.
One after-the-run structural receipt for a single agent trace. Best when you need proof before adding live monitoring.
One recurring agent workflow, zero-semantic receipts, and a dashboard for structural health review.
The recommended pilot for teams running agents often enough that loops, retries, and token exposure can repeat.
Higher-touch onboarding for multiple workflows and deeper review support.
CAUM observes structural events, counters, hashes, task family, status, and phases. It does not need prompts, files, source code, or customer messages.
Receipt is the entry artifact. Live is the recurring monitoring layer. Everything else supports those two paths.
These sample reports are the clearest way to understand the product: a PDF for human review and a vault for audit evidence.
Use CAUM Receipt for after-the-run evidence. Use CAUM Live when you want structural monitoring while the agent runs.