Anthropic has launched a public beta for self-hosted environments for its Claude Code coding agent, allowing organizations to execute cloud sessions on their own infrastructure rather than through Anthropic’s servers. This feature is available to customers on Claude Team and Enterprise plans, though it is not enabled by default. The self-hosted sessions can be initiated from various platforms including web, mobile, and desktop, and operate within the customer’s network, integrating seamlessly with internal services, toolchains, and security protocols.
This development is particularly advantageous for teams with specific network, compliance, or tooling needs that necessitate maintaining control over execution environments. Key data management practices are enforced: repository checkouts, build artifacts, secrets, and modified files remain within the organization’s proprietary systems, while requests for model inference, including prompts and responses, are sent to Anthropic’s API. Organizations involved in the preview program reported three primary benefits: direct access to internal services without public internet exposure, the ability to pre-install development tools for seamless session initiation, and enhanced compliance through the retention of source code within internal systems.
Senior Engineering Manager George Jacob from Faire emphasized that self-hosted environments facilitate the integration of Claude Code into existing workflows while ensuring security and operational control. This implementation allows Claude to generate pull requests, address CI issues, and respond to developer activities with scalable compute resources. Notably, self-hosted environments differ from the Remote Control feature, as the latter allows users to continue local sessions remotely but lacks the shared infrastructure aspect provided by self-hosting.
On August 5, 2026, Anthropic also made notable enterprise moves, including enhancements to other features.