lend lets remote coding agents use laptop tools with approval
AI summary
The author introduced lend, a macOS and Linux tool that lets remote coding agents invoke laptop tools only after per-request user approval, while filtering SSH keys and keeping secrets out of the remote agent. It also supports a stable forwarded socket across reconnects, action records, and local approval prompts, including Touch ID on macOS. The author notes that approved commands run unsandboxed, VM-provided caller details can be forged, and interactive programs do not work without PTY support.
In-depth insight
lend offers a practical approval and credential-isolation layer for remote coding-agent workflows, though it does not sandbox commands that users approve.