Running coding agents without handing them my credentials
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
The developer Antonio released lend, a tool that lets remote coding agents use credentials on a local laptop only with explicit user approval, addressing the risk of copying tokens onto disposable VMs or forwarding an SSH agent that exposes every key without prompting. The tool fits into a workflow built around herdr, a terminal multiplexer for coding agents that tracks agent status and lets agents drive each other, and Ovis, a browser and phone client that surfaces agent approval requests as answerable cards with push notifications over Tailscale. The problem lend targets is that agents running on throwaway VMs still need to act as the user for tasks like monitoring CI with gh, calling APIs with 1Password tokens, and SSHing into other hosts. Copying credentials into files like ~/.config/gh/hosts.yml exposes them to every process on the machine for the token's lifetime, while ssh -A lets the VM use every key in the agent without asking. Lend keeps the socket stable across reconnects, hides secrets from the agent, and issues leases so users can see what happened. What to watch next is how much of the credential surface lend can cover, since the post itself flags protections it does not provide.
Muhabbit's AI wrote this summary in its own words from Hacker News's report. It can get details wrong, so check the source.
Named in this story
lend, herdr, Ovis, Codex, Claude Code