A unified portal to assess compliance, explore controls, and deploy your AI Gateway.
Sample figures. Connect in Step 3 to see your own live data here.
3,687 attempts to send credentials, source code, financials, and PII to external models were blocked. Unmonitored endpoints are a compliance risk.
Guardrails blocked 1,182 unsafe prompts before they reached the model: jailbreaks (558), PII (376), hate speech (133), violence (97).
The support copilot served 37.7% of requests from cache, saving $890. The Returns team hit 95% of its budget, triggering an alert.
Dynamic Routing splits traffic on usergroup, with a weekly budget and automatic fallback.
Via the evaluations API (patchLog()), apps capture thumbs up/down and model-as-a-judge scores (0–100) on gateway logs, so we can track prompt drift and response quality.
Every tool an AI agent calls is logged in the MCP Server Portal, attributed to a user identity. Outbound agentic requests are scanned by Cloudflare Gateway and DLP to prevent exfiltration.
Only 2 of 7 sanctioned AI apps have all controls active. 2 High-risk apps (incl. the HR resume screener) run with logging, auth, and DLP off.
| Application | Risk | DLP | Guardrails | Auth | Logging |
|---|---|---|---|---|---|
| northwind-support-copilot | Limited | ✓ | ✓ | ✓ | ✓ |
| checkout-rag-assistant | Limited | ✓ | ✓ | ✓ | ✓ |
| hr-resume-screener | High | ✗ | ✓ | ✗ | ✗ |
A guided tour of the real AI Gateway interface: logs, analytics, guardrails, and dynamic routing.
Deploy a live AI Gateway (ai-control-plane-demo) on your account, with budgets, caching, and safety rules pre-configured.
We verify it, resolve your account, and unlock deployment.
Rather not paste a key here? It's open source, run it yourself locally and your token never leaves your machine.
Creates a gateway with caching, budget rate limits, and prompt logging, live immediately.
Flip Step 1 between the sample story and your live account.