Puckyto

Puckyto

**A mission control for your AI coding agents.**


Puckyto is a native macOS app that runs several terminals side by side and treats the AI agent inside each one (Claude / ChatGPT / Gemini) as a first-class citizen: it has a name, a task, a rule set and a memory that survives restarts. Around them sit the things you actually need when three agents are working at once — a shared board they write to, a task queue that feeds them, git checkpoints that make their edits reversible, real token accounting, and a live map of who is doing what.

It does not wrap or replace the agent CLIs you already use. It owns their terminals, gives them context, and gets out of the way.

**Less about keeping sessions alive, more about giving each agent an identity, context and measurable output.**

*[Türkçe README](README.tr.md)* · 📘 **[Setup & Build](docs/SETUP.md)** · 📗 **[User Guide](docs/USAGE.md)**


./scripts/build-app.sh && open "build/Puckyto.app"

Requirements: macOS 14+, Swift 5.9+, and whichever agent CLI you want to run — [Claude Code](https://claude.com/claude-code) (`claude`), OpenAI Codex (`codex`) or Gemini CLI (`gemini`). The interface is English by default, Turkish is one click away.

---

Why it exists

Running one coding agent is easy. Running four is a different job: you lose track of which one is blocked, they overwrite each other in the same folder, you re-type the same prompts, and you have no idea what the day cost you. Puckyto is built for that second situation.

The problem	What Puckyto does
"Which agent is waiting for me?"	Status chips, a menu-bar summary and notifications you can click to jump straight to that terminal
"What did they change?"	A git checkpoint per agent start, a live Δ +a −b chip, one-click revert
"They keep colliding"	A same-fold