Integrating Workers AI and AI Gateway into a Single AI Control Plane

Cloudflare is integrating Workers AI and AI Gateway into a single AI control plane. Users can connect to all model providers, including Workers AI, via the same path, and manage observability, logging, security, access control, and billing in one place.

Workers bindings and the REST API serve as the unified entry point. By specifying gateway.id in env.AI.run() calls, or by using the /ai/ REST endpoint with the cf-aig-gateway-id header, Workers AI requests can be processed through AI Gateway.

Specifying the default gateway automatically creates it on the first authenticated request without additional configuration. Subsequently, the following information can be reviewed for all requests:

• Full request and response payloads
• Token usage by model
• Cost attribution
• Latency analysis
• Error rates

Existing direct invocation methods can be modified by adding gateway: { id: 'default' } as the third argument. If per-application traffic isolation or custom caching rules are required, you can switch to named gateways.

Additionally, integrated billing is supported to allow AI Gateway credits to be applied to Workers AI usage. Previously, AI Gateway credits could only be used for external model providers such as OpenAI and Anthropic.