Hey Product Hunt, I work with coding agents all day, and what actually slows me down is not the agent: it's me. One agent per project, five projects open, and I am the one polling terminals to find out which one finished, which one is stuck waiting on an answer, and which one has been idle for twenty minutes because I did not notice. Voyageur is the control tower I wanted for that. Three things, and deliberately only three: 1. Fan-out. Launch and follow many agents across many projects from one window, on your laptop or on remote machines. 2. An attention router. Instead of you scanning, it surfaces the agent that just finished or is waiting on you. 3. Out-of-band signal. It reaches you when the screen is shut, so "start it and walk away" actually works. On agents: Claude Code is supported natively today. Voyageur reads its real state, down to which subagents are running, instead of scraping a terminal and guessing. That is precisely what makes the attention router able to tell "still thinking" from "waiting for you". Other agent CLIs are on the roadmap. What will not change: Voyageur does not resell inference. You bring your own Claude Code subscription and you keep root on your own machines. Running a workspace on our infrastructure is there for convenience, never the point, and never required. The alpha is open now and runs to 31 August. It is open to everyone and everything is unlimited during it. macOS (Apple Silicon) and Linux are available today. Windows is coming through WSL2, on the VS Code Remote model rather than a degraded cmd/PowerShell port, and there is a waitlist on the download page so I can size the real demand instead of guessing. After the alpha the control tower stays free. Three comfort features become paid: the hosted environment, fan-out beyond two parallel workspaces, and centralised skills backup. Alpha testers who send usable feedback keep unlimited local fan-out for life, 50% off hosting for as long as the subscription runs, and the first hosted month free. Cloud seats are capped at 50 for capacity reasons and I approve them by hand. It is an alpha. It is genuinely usable, I run it every day, and it will have rough edges. Break it and tell me. One thing I would honestly like to know: how many agents do you run in parallel today, and what makes you stop adding more? If the answer is the tracking overhead rather than the cost, that is exactly the thing I am trying to kill. Steven

Voyageur

Launch your coding agents. Close your laptop.

Launch your coding agents. Close your laptop.