Claude Code Can Now Run on Your Own Infrastructure, Not Just Anthropic's

Anthropic announced on August 6, 2026 that Team and Enterprise customers can now run Claude Code sessions on infrastructure they fully control themselves, instead of exclusively through Anthropic's own hosted environment. The change gives those customers internal network access the hosted version could not reach, the ability to add custom tooling around the agent's execution environment, and their own compliance controls over how and where the agent actually runs.

The feature ships off by default. A Team or Enterprise admin has to explicitly enable it, which keeps the hosted environment as the default path for customers who have not made a deliberate decision to change it.

The practical unlock is for teams that could not adopt Claude Code at all under a hosted-only model, because internal network access or specific compliance requirements made a third-party hosted execution environment a non-starter. Self-hosting the execution layer does not change what the model can do, but it changes what it can reach, and for a coding agent, what it can reach is often the actual blocker to using it on real, non-public codebases.