Running three Claude Code sessions across three terminal tabs while juggling a Codex process in a fourth — that’s the current state of multi-agent coding. Herdr is an open-source CLI tool that puts all of them in one place.
The pitch is simple: a terminal multiplexer built specifically for AI coding agents. Instead of alt-tabbing between agent sessions, Herdr lets you launch, monitor, and coordinate multiple agents from a single interface.
What it does concretely:
- Spawn multiple agents — run Claude Code, Codex, or other CLI-based agents as parallel sessions within one terminal
- Live monitoring — watch each agent’s output in real time without switching windows
- Coordination layer — route tasks to specific agents and track what each one is working on
- Terminal-native — no Electron app, no browser UI, just your shell
The project picked up 77 points and 47 comments on Hacker News, which suggests the pain point resonates. Developers who’ve adopted agent-heavy workflows — running one agent on frontend, another on backend, a third writing tests — currently rely on tmux or multiple terminal windows with no awareness of each other. Herdr adds a purpose-built abstraction on top.
This sits in a growing category of “agent orchestration” tools that treat AI coding agents less like chat interfaces and more like managed processes. The difference from something like a VS Code multi-agent extension: Herdr stays in the terminal, which matters for developers who already live in tmux/Neovim workflows and don’t want to context-switch into an IDE.
NOTE
Herdr is early-stage and open source on GitHub. The agent ecosystem it targets — CLI-based tools like Claude Code and OpenAI Codex CLI — is itself only a few months old, so expect the interface to shift as those upstream tools change.
The repo is live at github.com/ogulcancelik/herdr. Worth watching if you’re already running more than one agent session at a time and tired of the tab sprawl.