Back to Feed

Claude Code Introduces Two New Sandbox Features for Enhanced Security

Claude Code now features two new sandboxing options that enable developers to work more securely by reducing permission prompts and increasing safety. The new features are available as a beta research preview and have been open-sourced for others to adopt.
Making Claude Code more secure and autonomous with sandboxing - Anthropic
www.anthropic.com
Claude Code Importance: 7 positive

Fyra's Brief:

Claude Code now includes a sandboxed bash tool that allows developers to run commands within defined limits without permission prompts, improving safety and productivity.
The new sandboxing features, built on top of operating system-level primitives, provide filesystem and network isolation to prevent prompt-injected Claude from modifying sensitive system files or leaking information.
Claude Code on the web enables users to run Claude Code in an isolated sandbox in the cloud, ensuring sensitive credentials remain secure even if the code running in the sandbox is compromised.
The new features are available as a beta research preview and have been open-sourced for others to adopt, aiming to improve the security posture of developers' agents.