Peaky Panes

████    █████    ███    █   █   █   █    ████      ███    █   █    █████    ████
█████   ████    █████   ████     ███     █████    █████   ███ █    ████    ████ 
█       █████   █   █   █  ██     █      █        █   █   █  ██    █████   █████

Multi-project orchestration in a single TUI for AI agents and dev environments.

Run all your projects and AI agents in one dashboard instead of juggling terminal windows or tabs. Define layouts in YAML, share them with your team via git, and get consistent environments everywhere. Sessions are owned by a native daemon so they keep running after the UI exits.

Why?

Running multiple agents on different projects at the same time takes real organization and a different approach to how we do agentic coding.

I built Peky to solve my own problems with juggling too many windows and wasting time on setup and rebuilding the same environment every time. It lets you build your own reusable window layouts and better-organized projects.

Features

• 🧠 AI agent orchestration - Run Codex/Claude/agents side by side with quick replies, slash commands, and broadcast
• 🗂️ Multi-project dashboard - See every project/session in one TUI without switching windows
• 📦 Built-in layouts - Works out of the box with sensible defaults
• 🧩 Exact grids - Use grid: 2x3 for consistent rows/columns
• 📁 Project-local config - Commit .peakypanes.yml to git for team sharing
• 🏠 Global config - Define layouts once, use everywhere
• 🔄 Variable expansion - Use ${EDITOR}, ${PROJECT_PATH}, etc.
• 🎯 Zero config - Just run peky in any directory
• 🧠 Native live previews - Full TUI support (vim/htop) with live panes
• 🧭 Persistent native daemon - Sessions keep running after the UI exits
• 📜 Scrollback + copy mode - Navigate output and yank from native panes
• ⌘ Command palette - Quick actions, including renaming sessions/panes
• 🖱️ Mouse support - Click to select panes, double-click to focus a pane

Quick Start

Install

Using npm (recommended)

npm i -g peakypanes
peky

Run once with npx

npx -y peakypanes

Using Homebrew

brew tap regenrek/tap
brew install regenrek/tap/peakypanes
brew services start peakypanes
peakypanes

Using Go

go install github.com/regenrek/peakypanes/cmd/peky@latest

Usage

Start peakypanes:

peky start

Custom Configuraton

Create project-local config (recommended for teams):

cd your-project
peky init --local
# Edit .peakypanes.yml
git add .peakypanes.yml  # Share with team

The CLI is peky (alias: peakypanes).

Documentation

• Configuration & layouts
• Layout builder guide
• Dashboard & keybindings
• CLI reference
• Performance tuning
• Testing
• Daemon & troubleshooting
• Release process

Configuration (quick look)

Project-local config lives in .peakypanes.yml, and global config lives in ~/.config/peakypanes/config.yml. See the full schema, built-in layouts, and variable expansion in docs/configuration.md.

session: my-project

layout:
  panes:
    - title: editor
      cmd: "${EDITOR:-}"
    - title: server
      cmd: "npm run dev"
      split: horizontal

Dashboard

The dashboard shows projects, sessions, live previews, and a quick-reply bar for slash commands and broadcasts. When the quick reply input is empty, common terminal keys (Enter/Esc/arrows/tab) are forwarded to the selected pane for interactive prompts. See docs/dashboard.md for keybindings, slash command behavior, and dashboard configuration.

CLI

See docs/cli.md for the full command list and examples.

Testing

See docs/testing.md for the full test matrix and CI notes. For heavy concurrency/regression coverage, run scripts/cli-stress.sh locally. For profiling + benchmarks, see docs/performance.md.

Release

See RELEASE-DOCS.md for the full release checklist (tests, tag, GoReleaser, npm publish).

Windows

npm packages are currently published for macOS and Linux.
Windows users should install from the GitHub release or build with Go.

For Teams

1. Run peky init --local in your project
2. Customize .peakypanes.yml for your stack
3. Commit to git
4. Teammates install peakypanes and run peky (alias: peakypanes) - done!

License

MIT

Credits

@pi-mono - Peky Agent is inspired by PI / Mario Zechner's work. (Direct pi integration would've been ideal, but 60 MB of Node for agent + sidecar isn't worth it) @charm - peaky is using bubbletea and lipgloss

Links

• X/Twitter: @kregenrek
• Bluesky: @kevinkern.dev

Courses

• Learn Cursor AI: Ultimate Cursor Course
• Learn to build software with AI: AI Builder Hub

See my other projects:

• codefetch - Turn code into Markdown for LLMs with one simple terminal command
• instructa - Instructa Projects