Anthropic Launches Claude Apps Gateway for Enterprise Controls Across Bedrock and Google Cloud

A control layer for Claude applications

Anthropic has introduced Claude Apps Gateway, a self-hosted gateway designed to help organizations manage deployments of Claude Code and Claude Desktop. In its announcement, Anthropic said the gateway sits between Claude applications and supported inference providers, including Amazon Bedrock, Google Cloud, and Anthropic's own platform.

The release is aimed at companies that want to provide access to AI tools while maintaining centralized oversight of authentication, usage policies, operational data, and spending. Rather than replacing the underlying model service, Claude Apps Gateway acts as an intermediary through which approved applications connect to Claude inference endpoints.

Centralized identity, policy and observability

According to Anthropic, the gateway brings together single sign-on, policy controls, per-user cost attribution, OpenTelemetry Protocol (OTLP) telemetry, multi-upstream inference routing, and spending caps.

Single sign-on is intended to connect Claude application access with an organization's existing identity systems. Policy controls can give administrators a central place to define permitted access patterns, while telemetry can help operations teams monitor requests and investigate service issues.

Per-user attribution and spending caps address a separate enterprise concern: controlling the cost of broadly deployed AI tools. These features can allow organizations to associate usage with individual users or groups and set limits intended to reduce the risk of unplanned consumption.

InfoQ described the product as a self-hosted control plane for Claude Code and Claude Desktop, highlighting the gateway's identity, policy, telemetry, routing and spend-control features. The publication also noted that Anthropic, AWS and Google Cloud have released deployment guidance for the offering.

Routing between inference endpoints

Multi-upstream routing is a central part of the gateway's design. Anthropic says organizations can use it to direct application requests across supported inference back ends, giving them a management point above the individual model provider endpoint.

AWS's deployment guidance focuses on running Claude Apps Gateway in AWS environments. AWS said the gateway can be deployed on Amazon ECS, Amazon EKS, or Amazon EC2, using PostgreSQL for data storage. Under that configuration, it can route requests either to Amazon Bedrock or to Anthropic's Claude platform on AWS.

AWS also describes regional and cross-account failover options. For organizations operating across multiple AWS accounts or regions, those options could support internal availability planning while keeping application access and governance controls in one place.

An operational focus for enterprise AI

The launch does not introduce a new Claude model. Instead, it addresses the operational requirements that emerge when AI coding and desktop applications move from limited tests to wider organizational use.

For enterprises, model selection is only one part of deployment. They also need ways to connect tools to existing identity systems, enforce internal rules, observe usage, allocate costs, and manage continuity across approved cloud services. Claude Apps Gateway is Anthropic's proposed layer for those functions in Claude Code and Claude Desktop environments.

Its practical value will depend on how readily organizations can integrate it with their identity providers, cloud deployments, PostgreSQL operations, and observability systems. The AWS guidance provides one implementation path, while Anthropic's announcement positions the gateway for use across Amazon Bedrock, Google Cloud, and Anthropic-hosted Claude services.