Cloudflare Identity-Aware AI Gateway: Every AI Request Now Tied to a Real Employee

How it ties employee identities to AI requests

The gateway sits behind Cloudflare Access. Before any request reaches an AI model provider, the user must authenticate against the organization's existing identity provider (IdP) . When the request arrives through an Access-protected custom domain, Cloudflare validates the user and device posture and writes the verified Access user ID into request metadata as cf.user_id .

This replaces blind, shared API keys that made it impossible to attribute requests to specific people or systems . IT teams now know exactly who prompted which model, at what time, and from which device.

Visibility and cost control

Once identities are attached, the gateway unlocks granular cost management:

• Per-user spend limits — IT teams can set spending caps at the individual or team level, not just account-wide .
• User Insights — A companion feature that learns normal AI usage patterns per user or automated system, then alerts when activity or spending deviates from that baseline .
• Log and analytics filtering by user — Requests can be filtered, audited, and analyzed by authenticated identity in logs and analytics dashboards .
• Real-time flagging — The system can detect when employees use expensive models for simple tasks that cheaper alternatives could handle .
• Cost-saving features — Repeat requests can be cached to cut redundant costs; rate limits stop runaway usage before it hits the invoice .

Companion security and governance features

Identity alone isn't enough — enterprises also need guardrails:

• Data Loss Prevention (DLP) — Automatic filters strip employee names, passwords, and other sensitive data out of requests before they reach external AI providers .
• Guardrails — Content filtering and policy enforcement at the gateway layer for both input prompts and model responses .
• Rate limiting — Per-user or per-gateway request throttling to prevent abuse .
• Dynamic routing — A beta feature that allows requests to be routed to different models based on policy .

Future and related capabilities

Cloudflare is extending identity-aware controls to autonomous agents:

• Cloudflare Wallets & cloudflare.pay — Announced August 4, 2026, these give AI agents a stable identity and the ability to make purchases within limits set by their human creators .
• Code Mode — Launched during Agents Week (July 2026), designed to slash token costs for coding agents .
• Shadow MCP detection — Cloudflare Gateway rules to detect unauthorized Model Context Protocol servers .

The service is available now in open beta. As Dane Knecht, Cloudflare's CTO, summarized: "Teams get the freedom to work with any AI model they choose. IT gets the real-time visibility, guardrails, and budget controls they actually need" .