INFRA Signal 583
Unifying Workers AI and AI Gateway into a single AI control plane
↗
Cloudflare merged Workers AI and AI Gateway into a single AI control plane, consolidating bindings, APIs, observability, and billing.
Developers can now route requests to any model, whether hosted on Cloudflare GPUs or external providers, through one unified endpoint, gaining automatic logging, token accounting, and cost attribution without extra setup. The change also lets users spend AI Gateway credits on Workers AI and unlock higher rate limits, simplifying budgeting across providers. Existing code that called Workers AI directly still runs, but it will miss the new unified telemetry unless updated.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
A single binding and /ai/ REST endpoint now handle both Workers AI and external model calls, removing the need to choose between products.
Specifying "default" as the gateway auto-creates an AI Gateway on first use, instantly providing full request/response logs, latency metrics, and token usage tracking.
AI Gateway credits can be applied to Workers AI usage, and unified billing offers higher rate limits for workloads that adopt the new path.
THE READ
What the cluster adds up to.
Previously, developers had to decide between two separate services: Workers AI for Cloudflare-hosted models and AI Gateway for proxying to external providers. Each product required its own binding or API call, and observability was only available after manually provisioning a gateway. The unification collapses these distinctions, presenting a single call path that can target any model while automatically attaching the control plane features. To adopt the new flow, code must be updated to include an optional gateway argument or to use the new /ai/ endpoint. For most use cases, passing the identifier "default" is sufficient and triggers automatic gateway creation, meaning no additional configuration is required. Teams that need custom routing, caching, or separate billing can still define named gateways, but this adds a small management step. The biggest operational benefit is the immediate availability of detailed telemetry: every inference request is logged with full payloads, token counts, latency breakdowns, and error rates. This data appears in the AI Gateway dashboard without any extra setup, enabling faster debugging and compliance auditing. The unified billing model
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER