Deploying Anthropic Claude Apps Gateway for AWS to Manage Enterprise AI Workloads

The Anthropic Claude apps gateway offers a centralized governance layer for deploying Claude Code and Claude Desktop across enterprise environments on AWS. It addresses key operational needs such as authentication, model access control, cost management, and telemetry, streamlining governance at scale.

The deployment architecture involves running the gateway within AWS Fargate inside a Virtual Private Cloud (VPC), with options for Amazon EKS or EC2. It utilizes components like Amazon RDS for session and spend data, an internal Application Load Balancer for TLS termination, and VPC endpoints for secure service connectivity. Authentication is managed via OIDC providers, supporting integrations with providers like Okta, Microsoft Entra ID, and Amazon Cognito.

Request flow begins with a developer signing in through their identity provider, obtaining a short-lived token, which is then used for inference requests. The gateway validates tokens, enforces policies, and routes requests to Amazon Bedrock or Claude Platform on AWS, with telemetry data relayed via OpenTelemetry Protocol.

The system enforces governance through features such as identity management, centralized model access policies, per-user telemetry, inference routing with failover, and spend caps. Policies are defined in YAML, scoped by user groups, and enforced server-side, ensuring compliance and operational control.

Deployment patterns vary from single-team, single-region setups to multi-account architectures, allowing organizations to tailor the deployment to their structure and needs. The gateway supports failover across regions and providers, and can be integrated with organizational cost management tools for inline spend enforcement.

The article concludes with guidance on setup, including cloning the GitHub repository, choosing deployment options, and consulting the detailed documentation for configuration and operational management.