In a coordinated multi-cloud expansion, Anthropic has officially launched the Claude Apps Gateway for Amazon Bedrock and Google Cloud. The announcement coincided with Claude models becoming generally available on Microsoft Foundry via Azure, effectively establishing a unified enterprise control plane across all three major hyperscalers.

The Claude Apps Gateway functions as a seamless, plug-and-play proxy designed to solve a growing enterprise headache: letting hundreds of developers use agentic tools like Claude Code and custom workplace workflows without managing fragmented API keys, risking security leaks, or running up unmonitored token bills.

Diagram showing Claude Code clients connecting over HTTPS with bearer tokens to a self-hosted Claude apps gateway inside your infrastructure, which signs users in against your IdP, stores auth state in PostgreSQL, relays telemetry to your OTLP collector, and forwards inference to Amazon Bedrock, Google Cloud, Microsoft Foundry, or the Anthropic API

1. One Variable to Rule Multi-Cloud Routing

The core mechanism of the Apps Gateway relies on a single configuration hook. Because Claude Code and the Anthropic SDK natively look for an environment variable called ANTHROPIC_BASE_URL before sending data, the gateway intercepts this traffic completely transparently.

Instead of routing code and prompts directly to Anthropic’s public servers, requests flow through the enterprise-controlled gateway, which dynamically maps the workload to the preferred cloud backend:

 [ Claude Code CLI / App ] ──► Points to ANTHROPIC_BASE_URL (Enterprise Gateway)
                                         │
                                         ▼ (Policy & Governance Inspection)
                     ┌───────────────────┼───────────────────┐
                     ▼                   ▼                   ▼
            [ Amazon Bedrock ]   [ Google Vertex AI ]   [ Microsoft Azure ]

2. Core Architecture: Solving the Enterprise Token Friction

As organizations scale their use of autonomous AI coding agents, standard API setups quickly break down. The Apps Gateway introduces native infrastructure controls to address these operational vulnerabilities:

  • Identity Mapping & Audit Trails: The gateway anchors both the human principal and the agent identity throughout the loop. If an engineer triggers an autonomous feature, the outbound request logs the human developer, while the subsequent model-driven tool execution is explicitly scoped to that specific conversation.
  • The Multi-Provider Emergency Brake: Administrators can set per-developer, per-team, or organization-wide budget ceilings. If a developer accidentally leaves an agentic loop running that burns high-context tokens, the gateway enforces hard cutoffs and automatically revokes access before costs spiral out of budget.
  • Protocol Dialect Translation: One of the steepest hurdles of multi-cloud AI has been syntax differences; for example, Google Cloud Vertex AI’s Claude endpoint speaks a slightly different dialect than Anthropic’s native API. The gateway acts as a real-time translator, allowing developer tools built strictly for the Anthropic SDK to interface perfectly with Vertex AI or Bedrock without any code changes.
  • Zero-Latency Data Compliance (PHI/PII): For highly regulated environments (such as healthcare deployers running under BAA or finance teams), the gateway runs localized token inspection passes. It can actively block or redact tool results before they return to the model if they contain unauthorized data, adding less than 50 milliseconds of latency overhead—well within standard streaming budgets.
Diagram showing Claude Code routing through a gateway. In a developer machines zone, the Claude Code CLI and VS Code extension send requests to the gateway address with a per-developer credential. In a zone labeled your infrastructure, the gateway handles authentication, usage tracking, budgets, and routing, and forwards requests with your organization's credential. In a model providers zone, a solid arrow leads to the provider you configure, shown as the Anthropic API, and dashed arrows lead to other provider options, illustrated with Amazon Bedrock, Google Cloud, and Microsoft Foundry as examples.

3. The Hyperscaler Footprint

By launching the gateway alongside its broader cloud expansion, Anthropic is anchoring its enterprise ecosystem directly into existing corporate agreements:

Cloud EnvironmentIntegration Depth & FeaturesInfrastructure Optimization
Microsoft FoundryGenerally Available natively on Azure. Works out-of-the-box with existing Azure identity, networking, and governance controls.Powered directly by Nvidia GB300 Blackwell Ultra GPUs to maximize prompt caching and extended thinking capabilities.
Amazon BedrockFully integrated via the Apps Gateway proxy. Maps client-side SDK requests directly onto AWS enterprise cloud pipelines.Allows organizations to leverage custom internal models or cheaper alternatives for routine tasks while reserving flagship tiers for heavy logic.
Google Cloud (Vertex AI)Supported seamlessly via the Gateway translation layer, binding API requests natively to GCP Application Default Credentials (ADC).Eradicates manual key management; developer authentication maps directly to pre-configured Google IAM policies and quotas.

The rollouts signify a clear transition into maturity for corporate AI adoption. By handing enterprise IT departments an architectural “kill switch,” automated budget enforcement, and strict data-handling guardrails, the Claude Apps Gateway shifts AI agents from rogue developer experiments into highly governed corporate infrastructure assets.