Summary
Anthropic launched a public beta of self-hosted environments for Claude Code on August 6, 2026, letting Team and Enterprise customers run agent sessions inside their own network on runners they operate, reaching internal services and data without exposing them to the public internet. A key caveat clarified in reporting: this is self-hosted execution, not self-hosted inference—prompts, tool results, and session transcripts still travel to Anthropic for inference and are retained so work can continue across surfaces.
What changed
August 6, 2026: Anthropic opened a public beta of Claude Code self-hosted environments for Team and Enterprise plans. Teams deploy long-lived runners (fixed or on-demand) that pick up sessions started from web, mobile, desktop, or routines and run each Claude Code process inside their own network. Reporting clarified the boundary: file checkouts, secrets, and build artifacts stay on customer infrastructure, but prompts, model responses, tool results, and transcripts are still sent to Anthropic for inference and retained.
Why it matters
Keeping agent execution on customer-controlled infrastructure removes a major blocker for regulated and security-sensitive teams, letting agents reach internal databases, registries, and toolchains without opening them to the internet. But because inference still runs on Anthropic and transcripts are retained, security and compliance teams must treat 'self-hosted' as execution locality, not data isolation—an important distinction for data-residency and vendor-exposure reviews.
Evidence excerpt
Self-hosted execution is not self-hosted inference: prompts, model responses, tool results, and session transcripts still travel to Anthropic for inference, and Anthropic retains the transcript so work can continue across web, desktop, mobile, terminal, and scheduled-session interfaces.