Run Claude Code sessions on your own compute
02:00 · August 6, 2026 · Claude Blog
Summary
Anthropic has opened a public beta for self-hosted environments that allow Claude Code sessions to execute on infrastructure controlled by the customer rather than on Anthropic-hosted systems. Sessions can be initiated from the web interface, mobile or desktop applications, or automated routines, after which they run inside the organization’s network alongside its existing services, toolchains, and security controls.
Repository checkouts, build artifacts, secrets, and any files created or modified during a session remain on the customer’s infrastructure. Prompts, responses, and tool outputs are still transmitted to Anthropic for model inference, and session transcripts are stored to support resumption across devices. The feature is available to organizations on Claude Team and Enterprise plans, is disabled by default, and is not offered to customers using zero-data-retention configurations.
To operate the environments, teams deploy a set of long-lived runner processes that accept sessions and launch a dedicated Claude Code instance for each one. Each session executes in an isolated checkout, allowing multiple concurrent sessions while keeping work separated between users and accounts. The setup differs from Remote Control, which ties a session to a single developer’s machine; self-hosted runners are managed by a platform or developer-productivity team and can serve any authorized user.
Anthropic notes that most enterprises will find the fully hosted offering simpler to operate and recommends self-hosting only when network, tooling, or compliance constraints require execution on customer-controlled infrastructure. Organizations choosing this route must allocate engineering resources to build and maintain runner images, keep runners updated, and manage the orchestrator when using on-demand mode.
Why it matters
This update is highly relevant for Dutch product teams and builders dealing with strict GDPR and data sovereignty requirements. By allowing local execution of Claude Code, enterprises can maintain tighter security controls over their proprietary code and build artifacts while leveraging advanced AI capabilities.
