Launching today
Herd
Run a herd of AI coding agents from one window
5 followers
Run a herd of AI coding agents from one window
5 followers
A lightweight desktop agent IDE for vibe coding at scale. Orchestrate dozens of AI coding agents from one desktop app, monitor every task in real time, and ship in parallel. No complex configuration or setup, just create or connect to a repo and get to work.
Hey Product Hunt!
We built Herd because, like a lot of you, we had three or four Claude Code sessions open across as many terminal windows and figured there had to be a better way.
There are some great multi-agent IDEs out there already, but we felt they were too heavy, bolting on functionality the agents themselves are already great at. So we went the other direction: keep it thin, lean on the agents, and stay out of the way.
What's in the box:
Lightweight by default: no config, no setup, no plugins. Point it at a repo and go. Cross-platform (Windows, macOS, Linux) with auto-updates.
Real parallelism: every agent runs in its own git worktree, so six agents can refactor the same file at the same time without stomping on each other. When you like what one did, commit, merge, or open a PR straight from the Source Control panel, with AI-generated commit messages and PR descriptions.
Multi-agent, multi-CLI: works with Claude Code, Codex, Kimi, and more side-by-side. Spawn agents per-workspace or as "free agents" with no repo at all (great for one-off questions or research).
Built-in dev surface: Run panel auto-detects your project and allocates ports, plus a real terminal, file browser with drag-to-chat, and inline diff viewer with VS Code-style preview/pinned tabs.
GStack-native: we used GStack heavily building this, so it's a first-class citizen. Dedicated panel, and active skills are surfaced throughout the app as agents use them.
It's free, it's local, and there's no account to sign up for. Would love your feedback, bug reports, and feature requests,
we're around all day.