Running coding agents without handing them my credentials
lend proxies a forwarded SSH-agent socket so a remote coding agent can request policy-checked commands executed on the credential-holding laptop, with local approval, instead of receiving the credentials themselves. It also supports scoped SSH signing and command rules.
This doesn’t make approved actions safe: HN commenters point out that a sufficiently powerful authorized operation can still leak data or create durable access, and discuss narrower permissions and sandboxing as alternatives.