Anthropic's Claude Code CLI tool reported to send user email addresses to external websites

On November 6, 2026, at 04:03, DEV Community published an article introducing an experimental release from Anthropic: sandbox-runtime, a lightweight CLI tool abbreviated as srt, that isolates community MCP servers used by Claude Code without requiring a Docker container. According to its README, the tool relies on native OS sandboxing — sandbox-exec on macOS, bubblewrap on Linux, and a dedicated local user account with network filtering on Windows — combined with proxy-based network filtering. It is designed to restrict file writes and network access, reducing the risk of local credentials such as SSH keys leaking. The article notes that a community MCP server runs with the user's full permissions, so a sloppy or malicious server added to Claude Code could read SSH keys or write anywhere the account can.

At 06:17 the same day, a post appeared on Hacker News titled that Anthropic's CLI tool Claude Code was found to be sending users' email addresses to external websites. Its summary says that, according to a bug report on GitHub, the development tool transmits users' email addresses to external websites, and that the security and privacy concern has drawn attention in open source communities, with discussion still spreading. The post links to issue 107757 in the anthropics/claude-code repository on GitHub, as well as a Hacker News discussion at news.ycombinator.com/item?id=50385272, and it gathered 111 points and 24 comments. At this point the matter appears to remain at the stage of community discussion: the available material does not indicate whether Anthropic has responded to the report or taken any follow-up action.