Demo. Sample data from a real deployment. Anything you change here lives in this browser session only and is not saved anywhere; sign out to start again.

Krynix Meridian Labs
Enforcement — inferred from observed decision evidence
Agent inventory

claude-code-platform

Connect with Claude Code

What governs it

Its own policies
coding-agent-guardrails@2
Org baselines
none apply to it

Every policy listed decides each call; the strictest verdict wins. Baselines are set for the whole org on the Policies screen.

Change its own policies

How it arrives

From the 6 most recent sessions it opened (79 events).

  • Claude Code — logs 69 events Proven against traffic captured from the real agent.
  • pdp 10 events Not an OpenTelemetry source Krynix describes field by field; the fields below are still measured from it.

Field by field

  • Signed evidence not checked here
  • Which agent ran arrived
  • Tool names arrived
  • Tool arguments arrived
  • Tool output not seen

    Claude Code — logs: This source does not carry it.

    Claude Code sends tool output on its traces, not its logs. Export traces to Krynix as well (`CLAUDE_CODE_ENHANCED_TELEMETRY_BETA=1` and `OTEL_TRACES_EXPORTER=otlp`).

  • Tool latency arrived
  • Tool failure rate arrived
  • Model arrived
  • Tokens arrived
  • Spend arrived
  • Errors not seen

    Claude Code — logs: Not measured for this source yet.

  • Prompts arrived
  • Plugins and hooks arrived