Launching today
nana
One terminal for every coding agent, on Windows
51 followers
One terminal for every coding agent, on Windows
51 followers
Nana is a native terminal multiplexer for Windows. Run Claude Code, Codex, OpenCode and more side by side, each on its own git worktree. See who is working, waiting or done, review diffs, and keep agents running after you close the app. Free beta.
Free
Launch Team / Built With
ClarifyThe Full-Stack AI CRM that replaces 5 tools
Promoted
Hi Product Hunt. I'm Gustavo, I built Nana.
I was running three or four coding agents at once and losing track of them across terminal windows. One would sit for ten minutes waiting for a "yes" while I was looking at another. So I built the terminal I wanted:
- Every agent in a split pane, each on its own git worktree, so they never touch each other's files.
- Nana recognizes 12 agents (Claude Code, Codex, OpenCode, Copilot, Cursor, Droid and more) and shows if each one is working, waiting for you, or done. No hooks, no setup. A Windows notification tells you when one needs you.
- Review diffs in the app and let the AI CLI you already use write the commit message.
- Terminals live in a small server process, so closing the window does not stop your agents. The same works over SSH on Linux and macOS.
It is a native Rust app (GPUI and Ghostty's terminal engine), not Electron. Free beta for Windows 10 and newer, no admin rights:
irm https://nana.areshk.dev/install.ps1 | iex
I'd love to hear what breaks and what you'd want next.
Doesn't it burn your computer? I used to use ZED with multiple tabs and it was making it very slow
@thomas_wazza It doesnt :)
an agent waiting for a yes in another terminal is an annoyingly easy thing to miss
the windows notification and worktree per pane are the two details that make this interesting to me
The native Rust choice is what sells this for me — GPUI plus Ghostty's terminal engine instead of another Electron shell should actually feel fast with that many panes open. And having the terminals live in a separate server process so closing the window doesn't kill running agents is exactly right; I've lost long agent runs that way before. Reviewing diffs in the same app instead of bouncing to another tool keeps the loop tight too.
Losing track of which agent is waiting on a yes is a real pain, the status view is a good call. When two agents on separate worktrees both touch a shared lockfile, does nana warn you before the merge, or is that left to git?