Cloudflare Unifies Workers AI and AI Gateway into Single Control Plane
AI model management now centralized in one platform.
1Evolution of Two Products
AI Gateway and Workers AI were initially designed as separate products. AI Gateway allowed users to send requests to any model provider with observability and security features. Workers AI provided API access to models hosted on Cloudflare's managed GPU infrastructure. User behavior showed these two products functionally converging over time, prompting the unification.
2Unified Control Plane
The new system enables users to manage all model providers (including Workers AI) from a single control panel. Billing, security policies, and usage analytics are centralized. This particularly benefits enterprise users who need to monitor multi-cloud AI services from one interface. Cloudflare aims to set a new standard in enterprise AI operations with this move.
3Automatic Observability
Integration eliminates the need for manual AI Gateway configuration. Using 'default' gateway ID activates observability features automatically on first request. Every request is logged, token usage tracked per model, and cost analyses generated automatically. These features provide quick start capabilities especially valuable for startups adopting AI solutions.
4Consolidated API Structure
The new /ai/ endpoint combines Workers AI and AI Gateway functionality under one REST API. Users no longer need to choose between products. API bindings are similarly unified to simplify developer experience. Cloudflare assures this change won't break existing integrations and will allow gradual migration.
5New Billing Model
Workers AI can now be paid using AI Gateway credits. Users can fund services across different providers like OpenAI, Anthropic or Workers AI from a single wallet. Unified billing users will receive higher rate limits on Workers AI. Cloudflare emphasizes this will enable cost optimization for organizations using multiple models.
6Model-First Routing
Coming soon, users will specify models rather than providers, with the system automatically selecting optimal providers and load balancing. For example, a request for Kimi K2.7 Code could be served by Workers AI, Moonshot API or another provider transparently. This feature will significantly reduce application downtime.
7Enhanced Resilience
Model-first routing enables automatic failover to alternative providers when issues arise, eliminating need for application-layer fallback logic. Cloudflare notes this capability is critical for zero-downtime applications. The system will also support requirements like Zero Data Retention while maintaining output quality through vetted providers.
8Industry Impact
Cloudflare's move could set a new standard in AI infrastructure management. This solution reduces operational complexity for multi-cloud organizations and may accelerate consolidation trends. By democratizing access to AI models, this approach could increase innovation velocity and enable next-generation applications across industries.