Industry story
Google Cloud and Anthropic launch enterprise gateway for Claude Code
cloud-costs engineering identity
The real news here is not that Claude Code got a gateway — it's that agentic coding just got a control plane enterprises can actually defend to their security team. Google Cloud and Anthropic's self-hosted Claude apps gateway handles SSO via OIDC and Google Workspace, enforces role-based access server-side, and tracks token spend against a Cloud SQL ledger with hard per-user caps. No more distributing API keys to developer laptops; onboarding and offboarding collapse to a single identity-provider group change. For any enterprise running Claude Code at scale in a regulated environment, the friction argument against agentic coding just got harder to make.
Full analysis
Google Cloud and Anthropic have announced the Claude apps gateway, a self-hosted service that sits between an organization's Claude Code clients (Anthropic's agentic coding tool) and Google Cloud's AI infrastructure. The gateway centralizes enterprise concerns that were previously left to individual developers: it handles identity via OIDC/Google Workspace single sign-on, enforces role-based access control policies server-side, attributes token usage to verified user identities for cost tracking, and enforces spend caps per user, group, or org against a Cloud SQL ledger — returning an error when limits are hit.
The gateway runs as a stateless container on Cloud Run and routes all inference through a single Cloud Run service account, keeping data, billing, and quota inside the customer's own Google Cloud project. This eliminates the need to distribute service-account keys or API keys to developer laptops, and makes onboarding/offboarding as simple as adding or removing a user from an identity provider group. The announcement is significant for agent builders deploying Claude Code at scale inside regulated or enterprise environments, as it provides a production-grade control plane for agentic coding workloads on Google Cloud.
Comments