AI Gateway Architecture on Google Cloud routes requests from multiple client types—Claude Code, Codex, IDE Assistants, and custom agents—through a Global HTTPS Load Balancer to three specialized Cloud Run services: Dashboard, LLM Gateway, and Control-Plane. The LLM Gateway abstracts access to diverse model providers including Vertex AI, Amazon Bedrock, Azure AI Foundry, StackIT, and Evroc, with secrets managed by Secret Manager and task orchestration via Cloud Tasks and Scheduler. Cloud SQL stores application state, while Cloud Logging and Cloud Monitoring provide audit trails and observability across the entire multi-cloud LLM ecosystem. This architecture decouples client applications from underlying model provider APIs, enabling seamless switching between LLM vendors and centralized governance. Fork this diagram on Diagrams.so to customize regions, add additional model providers, or adapt the control-plane logic for your organization's AI governance requirements.