alamops / agetor

What is agetor?

Agetor is a local kanban board for running Claude Code, Codex, and similar CLI coding agents in parallel. Instead of managing terminal windows, you create task cards, pick an agent, and watch output stream back through a web UI. Each task runs in its own git worktree, so two agents can hammer the same repo simultaneously without clobbering each other. Built in TypeScript on Bun with a React webview, it binds a localhost API and uses tmux to keep Claude sessions alive across multiple turns.

Why is it gaining traction?

The killer feature is worktree isolation combined with a visual control plane. Agents execute with your real shell privileges in real directories, but the kanban board gives you a dashboard for approvals, clarifying questions, and follow-up messages that would otherwise get buried in a TUI. You can run Claude interactively and send follow-up prompts without losing context. Everything stays local in SQLite - no cloud relay, no remote sandbox.

Who should use this?

Backend devs managing multiple repos who want to throw AI agents at a backlog without juggling tmux panes. Teams evaluating Claude Code or Codex at scale and wanting a way to track what each agent is doing across dozens of tasks. Anyone who has tried running two agents in the same repo and watched them overwrite each other's changes.

Verdict

Agetor solves a real orchestration problem with a clean local-first design, but at 13 stars it is early-stage software. The credibility score of 0.899% reflects a small, unproven codebase. The documentation is thorough and the test coverage is solid, but macOS is the primary platform and Linux/Windows builds are untested. Worth exploring if you need parallel agent orchestration today, but treat it as a hobby project with potential rather than production infrastructure.