By the end of this guide your team's AI usage will be intercepted, classified, and generating an audit trail your C3PAO assessor can verify.
One curl command pulls the Docker image, starts the proxy, and confirms it's running. Requires Docker Desktop.
curl -sSL https://houndshield.com/install | bash
Or non-interactive: HOUNDSHIELD_LICENSE_KEY=<key> UPSTREAM_API_KEY=<key> bash install.sh -y
Replace your AI provider's base URL with the proxy. Every SDK that speaks OpenAI-compatible APIs works — ChatGPT, Copilot, Claude, Gemini.
# .env or your shell profile OPENAI_BASE_URL=http://localhost:8080/v1 # ANTHROPIC_BASE_URL=http://localhost:8080/v1 # Claude # GOOGLE_BASE_URL=http://localhost:8080/v1 # Gemini
Verify the proxy is intercepting. This prompt contains a DoD contract number — it should be BLOCKED.
curl -X POST http://localhost:8080/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $OPENAI_API_KEY" \
-d '{
"model": "gpt-4o-mini",
"messages": [{"role":"user","content":"W911NF-23-C-0001 status?"}]
}'Expected response: { "error": { "code": "content_policy_violation", "message": "BLOCKED — CUI detected" } }
Download a PDF that documents every AI prompt your team sent, which CMMC controls were triggered, and your SPRS impact. Hand this to your C3PAO.
# From the dashboard — or via API: curl "https://houndshield.com/api/reports/generate?format=pdf&from=2026-04-01T00:00:00Z&to=2026-04-30T23:59:59Z" \ -H "Authorization: Bearer $HOUNDSHIELD_API_KEY" \ --output cmmc-audit-report.pdf
OPENAI_BASE_URL=http://localhost:8080/v1ANTHROPIC_BASE_URL=http://localhost:8080/v1GOOGLE_BASE_URL=http://localhost:8080/v1openai = OpenAI(base_url="http://localhost:8080/v1")llm = OpenAI(api_base='http://localhost:8080/v1')OPENROUTER_BASE_URL=http://localhost:8080/v1Your C3PAO assessment PDF is one click away in the dashboard. Every AI prompt your team sends is now scanned, logged, and audit-ready.