Hi Product Hunt, I'm Lohas, the maker of Agent OS. I built Agent OS because my AI work had become a pile of terminal windows. A task might start in Claude, continue in Codex, run on another machine, and end with a result I could not easily find a week later. The first issues from an actual user made the problem much clearer. They did not ask for another model. They asked for interrupted scheduled tasks to resume after an app restart, a way to steer an agent while it was running, timestamps and project folders tied to sessions, controls from message channels, and a safe way to move memory and task settings to another machine. Those requests changed Agent OS from a convenient launcher into an operating layer for agent work. Today it brings together: - structured sessions and native CLI terminals - local or authorized remote runtime hosts - tasks, schedules, run history, and deliverable review - side-by-side Web, CLI, and agent comparison - scoped memory and separate Markdown knowledge Agent OS is open source and available for macOS, Windows, and Linux. Sessions, memory, and knowledge stay on your machine by default. It does not replace Claude, Codex, or the other CLIs, and it does not bypass their accounts or permissions. I would especially value feedback from people who already use several AI coding tools. Where does your workflow still break? What would stop you from trusting Agent OS with recurring work?

Agent OS by AIUtil

Run, review, and remember work across AI coding agents

Run, review, and remember work across AI coding agents