Superagent

Driving a real browser on my own logged in sessions is the feature and also the part I'd think hardest about. An agent can't tell reading my inbox from sending from it, both are a tool call that returns success, and only one of them spends something I don't get back. The worktree per chat is what made me look twice, because the failure I actually hit isn't a bad diff, it's two agents editing the same file an hour apart with neither knowing. What I'd put on the page is the list of actions that stop and ask, since that list is the real product boundary.

the real browser on your own logged in sessions is the bit that will decide whether people keep this, and its also the scariest line in the whole post, which i mean as a compliment. the honest problem with watch it and take it back mid click is that watching doesnt scale. i watch the first ten minutes and then i stop, and the thing it does at minute forty is the one that actually mattered. my attention drops off and the permissions stay exactly where they were. so what id want sitting next to it is an allowlist on actions rather than on sites. reading anything is fine. clicking something that sends, posts, pays or deletes should come back to me, and that list is short enough to actually enumerate.

Hi Product Hunt 👋 I the maker of Superagent.Claude Code writes the code. Then it hands you a wall of scrolling text, forgets everything the moment you close the window, and leaves you alt tabbing to check whether the page actually changed.So I gave it a computer. A screen, a browser on your own logins, an iPhone in the window, your phone in your pocket. Built to feel like a Mac app rather than a terminal.✨ Beautifully built. Chats grouped the way you actually think. Panes and open files restored after a restart. A window that never raises itself when an agent finishes at 2am. Notifications that tell you what it actually said. Type mid turn and it reads you before it finishes.🖥️ A real browser. The one on your screen, in your own logged in sessions, which you watch and can take back mid click. Not a headless one you never see.📱 An iPhone living in the window. A real iOS Simulator streamed from its framebuffer. The agent taps and swipes it itself, so it builds an app and then checks it.🤝 Your phone in the loop. Pair the iOS app with a QR and approve what the agent asks from the sofa. End to end encrypted, through a relay that holds no key and stores nothing. Run your own with one command.🌿 A room per task. Start a chat on its own git worktree and it commits on its own branch. Your checkout never moves. Not one file.🔒 Private. No account, no server, no telemetry, no second subscription. It runs Claude Code locally on the plan you already pay for. All three pieces, the Mac app, the iPhone app and the relay, are MIT on GitHub. Do not trust me, read them.And anymore! I am here all day. Which part of the computer would you want next?