Alto and the tools you already use

If you already run coding agents, you probably have a way of doing it. One popular kind of tool is a desktop app that gives each task its own branch and worktree, with the diff and the pull request beside the chat. Another is a small terminal app that starts each agent in a tmux session. Some people use a terminal built for agents, with notifications and tabs that light up; some put tasks on a board and let an agent take each card; some send agents to run in a vendor’s cloud.

Alto isn’t one more of those, and it doesn’t ask you to give yours up. It’s the backplane underneath: the layer they all plug into. Those tools decide what the agents work on and how you review it. Alto takes care of where they run: terminals that stay alive, on every machine you have, that know when an agent needs you, and that any person, phone, script or agent can reach. Every one of those tools runs agents that are, in the end, programs in a terminal, and that’s exactly the layer Alto keeps.

Whatever you use, Alto adds the same things

• Sessions that are processes, not tabs. Each terminal is its own process, kept by a small service. Close the window, update the app, let something crash: the agent keeps working. See Sessions.
• Every machine, one address away. A session on another machine is studio:3, over Tailscale, with no ssh keys or open ports. Agents can run on the desktop under your desk, a build server, or a cloud VM, and you reach them all the same way. See Every session, every machine.
• Knowing what needs you. Working, needs you and done, for any program, exactly for agents that report it. Programs that already speak the open program status escape sequence (OSC 7501) show their status in Alto unchanged. See Agent status.
• Your phone. A notification when an agent needs you, and the terminal itself to answer it. See From your phone.
• A command line made for agents. Agents can read other sessions’ screens, run a command and get back only its output, wait for another agent to finish, and message the agents running elsewhere. alto guide tells them how. See The guide for agents.
• The terminal you like. None of this needs Alto’s app. See Keep your terminal.

Using it with what you have

A desktop app that gives each task a worktree. Keep using it to plan the work and review the diffs; that’s what it’s good at, and Alto doesn’t try to. Where you want an agent to outlive the app, run on another machine, or be answered from your phone, start it in an Alto session instead: from the app’s own terminal, alto new --attach -- claude is a session like any other. alto new --repo gives an agent its own checkout on a branch of its own, on whichever machine it runs, and alto fetch brings the branch back to review where you always do.

A terminal app that runs each agent in tmux. Run the manager itself in an Alto session (alto new --attach, then start it). It carries on as before, and now it outlives your terminal and can be reached from your other machines and your phone.

A terminal built for agents. Use it as your window onto Alto: alto attach studio:3 shows any session, from any machine, in whatever terminal you’re in. You keep the terminal you chose; the sessions just live somewhere sturdier, and Alto tells you (and your phone) which ones need you.

A runtime with its own server. You’ll find Alto’s ideas familiar: sessions that outlive the window, machines in one list. Alto’s take is a process per session, any terminal as a client, a command line agents use to reach each other, and your phone told when something needs you.

A board or issue tracker that hands work to agents. It decides what’s next; Alto is where the work then runs, and where you see it and step in. An agent picking up a card can start in an Alto session on whichever machine fits: alto new --in 'os:linux has:docker' --repo . picks one for you. See Putting work where it fits.

A cloud agent platform. If you’d rather your agents ran on machines you own, any server or VM becomes one of your Alto machines with alto install. See Servers.

What Alto leaves to the others

Alto is small on purpose. It doesn’t manage worktrees for you beyond --repo, review diffs, write pull requests, keep a task board, or plan the work. You and your agents decide the work, and the tools above do those parts well. Alto keeps the terminals they run in alive, everywhere, and tells you when one needs you.