GOVERN Probe

Every AI inference monitored. Zero application changes required.

Quickstart (30 seconds) Architecture overview

What is GOVERN Probe?

GOVERN Probe is a Docker container that sits between your application and any AI model API. It acts as a transparent proxy — your app sends requests to the Probe, the Probe forwards them to the model, intercepts the response, scores it across five dimensions, and streams telemetry to the GOVERN platform. Median added latency: under 4ms.

Five scoring dimensions

Dimension	What it catches
Security	PII exposure, prompt injection, credential leakage
Bias	Gender, racial, and age bias in model outputs
Accuracy	Hallucinations, ungrounded claims, factual drift
Behavioral Drift	Deviation from your established baseline
Cost	Token spend, budget burn rate, quota pressure

Three operating modes

Log — capture and report, never block. Zero risk to production traffic.
Flag — emit alerts when scores exceed thresholds. Human review triggered.
Block — hard stop on policy violations. Compliance-mode enforcement.

Deployment options

Docker run       → single container, any environment
Docker Compose   → sidecar alongside your app stack
Kubernetes       → DaemonSet or per-pod sidecar

Quick deploy

docker run -d \
  -p 4020:4020 \
  -e GOVERN_API_KEY=your_api_key \
  -e GOVERN_ORG_ID=your_org_id \
  -e UPSTREAM_URL=https://api.anthropic.com \
  archetypal/govern-probe:latest

Point your app at http://localhost:4020 instead of https://api.anthropic.com. Done.