Hey Product Hunt, I'm Tyler. I built Verlox alone over the past several months, because AI in the terminal has a trust problem that more intelligence doesn't fix. Agents are getting genuinely good at running commands. But when one deletes the wrong folder, no amount of intelligence brings it back. Every tool's answer is a permission prompt, and we all stop reading those by day two. Verlox is built on a different rule: never trust, always verify. You shouldn't have to believe the AI is right, because you can always see what it's about to do and always take back what it did. What that means concretely: It plans before it touches anything. Ask for something in plain English and Verlox writes out the full plan, every step scored for risk (reading files is low, installing is medium, deleting is high). Nothing runs until you approve. This is not a mode you toggle. It's the only way it works. Every command becomes a block. Output plus a plain English line saying what actually happened. Failed commands get a Fix this button that explains the error and hands you the corrected command. Successful ones get Explain, written by the model you picked, credited by name. Undo is real. Anything deleted is captured in the Recovery Vault first, restorable in one click. Every action lands on a timeline you can read like a story. "What did it do while I wasn't looking" always has an answer. Two surfaces. A real terminal (runs vim, REPLs, even other AI CLIs) for when you want your hands on the keys, and an AI terminal for when you'd rather just say what you want. It runs eleven hosted models plus one fully offline, and you can bring your own API keys so nothing locks you to my backend. Free to try, Pro is 15 dollars a month. Windows today. One question for you: what would an AI have to do in your terminal before no undo button could make you comfortable? Genuinely asking, because that answer is my roadmap. Tyler.

Verlox

The AI terminal that shows its work and can undo it

The AI terminal that shows its work and can undo it