Herdr is a desktop terminal manager built around a simple structural insight: terminals should outlive the windows that show them. Close the app, your shells keep running. Reopen, and your workspace is where you left it. This is tmux thinking applied to a native GUI, with a twist — Herdr also tracks coding agents across machines, surfacing which ones are blocked, done, or still grinding. The architecture splits cleanly: herdr-core runs your terminals and agents; herdr-app is the viewport. This separation means remote machines keep working when your laptop sleeps, and multiple screens can share the same workspace structure while keeping independent focus. It's the kind of design that rewards you once you have three machines and six agents running, and feels like overhead when you have one terminal open. The agent sidebar is the most distinctive feature. Agents report status — Working, Blocked, Done, Idle — and you can peek at any agent's terminal without disrupting your layout. Workers spawned by other agents show up automatically. For the growing cohort of developers running Cursor, Aider, or Claude Code across multiple repos simultaneously, this is solving a real pain point: the "which of my eight terminal tabs is screaming for input" problem. Mods are JavaScript programs that draw panels, menus, and dialogs — essentially a layout system you can script. Herdr ships built-in layouts including the delightfully named "Herdr 98." You can duplicate and modify shipped layouts; your copies persist through updates. This is smart — it avoids the VS Code extension ecosystem's fragility while still offering customization. But mods are unsandboxed trusted code, which the docs correctly flag as a security concern. The cross-platform story is solid: Linux (x86-64 and ARM64, needs Vulkan), macOS (Apple Silicon native, Intel from source), and Windows (x64, PowerShell installer). Built in Rust 1.95.0 and Zig 0.16.0, with a benchmarking harness that measures render-to-paint elapsed time across platforms. The project includes honest caveats about what the benchmarks do and don't measure — a good sign. Telemetry is on by default in release builds, off-able in settings. Both installers download and run code from the internet, and the docs invite you to read the scripts first. This transparency is welcome but the default-on telemetry will be a friction point for the privacy-conscious developer audience this targets. Herdr is making a bet that the near future of development involves managing fleets of coding agents across multiple machines, and that the terminal — not the IDE — is the right control surface for that. Whether that bet pays off depends on how quickly agent-heavy workflows become mainstream. For developers already deep in multi-agent setups, this is worth installing today. For everyone else, it's worth bookmarking.
