Introducing the Claude apps gateway for Amazon Bedrock and Google Cloud

Summary

The Claude apps gateway provides a self-hosted control plane that lets organizations manage developer access to Claude Code on Amazon Bedrock and Google Cloud without provisioning individual cloud credentials or manually distributing configuration files. Previously, teams had to maintain separate per-developer accounts, push settings to each workstation, and build custom tooling to track usage and costs. The gateway centralizes these functions through corporate single sign-on, role-based access controls, and per-user cost attribution while keeping inference traffic and telemetry within the customer’s infrastructure.

Deployed as a single stateless container on Linux and backed by a PostgreSQL database, the gateway stores upstream credentials, authenticates users against an existing identity provider, and distributes managed settings at login. Because the gateway logic ships inside the same claude binary developers already use, the client automatically applies policies on every request and supports consistent enforcement without additional client-side configuration. Onboarding reduces to adding a user in the identity provider; removal from the provider immediately revokes access.

The gateway reports usage data to a collector the organization operates and does not forward inference requests or telemetry to Anthropic unless explicitly configured to use the Claude API. Anthropic has also published the underlying protocol, allowing third-party implementations to support the same authentication and policy features. The result is enterprise-grade oversight of model access and spend on Bedrock and Google Cloud while preserving data residency requirements.

Why it matters

Directly actionable for Product Teams and Builders deploying Claude in regulated environments, with concrete steps for SSO, policy, observability, and cost management that align with EU data governance needs.