Claude Code v2.1.224: self-hosted runners and sandbox credential masking

Claude Code v2.1.224 lets you run Claude sessions on your own hardware. The claude self-hosted-runner command turns any machine or container into a host for web, mobile, and desktop sessions — available on Team and Enterprise plans.

The release ships six concrete changes:

• Self-hosted runners — claude self-hosted-runner exposes your own infrastructure as a compute target for Claude Code sessions across web, mobile, and desktop
• Archive plugin installs — install plugins from a zip over HTTPS, no git or npm required, with optional SHA-256 pinning for supply-chain safety
• Bedrock region prefix — ANTHROPIC_BEDROCK_REGION_PREFIX env var overrides the AWS_REGION-derived inference profile, letting you pin a specific cross-region profile
• Cross-session gating — crossSessionInbound and dialogExpiry settings hold messages sent to sessions with bypassed permissions for manual approval; messages to other sessions auto-deliver
• Sandbox credential masking — extract and onExtractNoMatch handle structured env values, decode: "jwt" with maskClaims does JWT-aware masking, and awsPairs / sigv4 re-sign AWS SigV4 requests; all require network.tlsTerminate
• Paste safety — removing an unavailable paste now triggers a cancel-and-confirm step before it modifies a command’s text

The self-hosted runner is the standout for teams with compliance constraints. If your org requires data residency or air-gapped execution, you can now keep Claude Code compute inside your own VPC instead of relying on Anthropic’s hosted infrastructure.

The sandbox masking options solve a narrower problem: when Claude Code intercepts TLS traffic for sandboxing, credentials in environment variables and JWTs can leak into logs. The new masking options redact those before they hit the output stream.

Available now in v2.1.224. Self-hosted runners require a Team or Enterprise plan.