Getting started
Your AI agents, organized in one view
Peky gives you a complete project overview with all your AI assistants working side by side. See your files, chat with multiple agents, and run commands—all in one place.
Quick start
Just open a terminal and type peky.
That's it—no configuration needed. Peky automatically detects your project and sets up your workspace.
Everything in one place
The project overview shows your file tree, multiple AI assistants (like Opencode and Claude), terminal output, and chat interfaces—all synchronized to your current project context.
Easy navigation
Use keyboard shortcuts to move between panes, open projects, toggle the sidebar,
and focus terminals. Press Ctrl+Shift+P for the menu.
Multiple AI agents
Run Codex, Claude, Opencode, and other AI assistants simultaneously. Each agent stays aware of your project files and can help with different tasks in parallel.
Why Peky
Everything you need to run parallel agents without losing context
Keep every session, pane, and command wired to the project they belong to, with a single app that stays running even after the UI exits.
AI agent orchestration
Run Codex, Claude, pi, and opencode side by side with quick replies and broadcast.
Multi-project dashboard
See every session in one place with live previews and instant switching.
Native daemon sessions
Sessions persist after the UI exits, so agents keep working in the background.
Zero config startup
Run peky in any repo and it just works, no setup required.
Project-local layouts
Commit .peky.yml for team-shared environments that stay consistent.
Built for power users
Command palette, scrollback, copy mode, and TUI-friendly live panes.
How it works
One layout, many agents, total clarity
Build reusable layouts, then launch every project with a dashboard that keeps context and actions in a single view.
- 1. Open a terminal (Ghostty preferred)
- 2. Install with
npm i -g peky - 3. Run
pekyin any directory
Ready to organize your AI agents?
Peky is open source and free to use. Star the repo, contribute, or just get started.
View on GitHub