jebi is a terminal emulator for Mac with built-in local AI. It explains failed commands, suggests what to run next, and lets you ask questions about your session — all without an API key or internet connection.

• Error explanations — when a command fails, jebi explains why and how to fix it
• Natural language to command — press ⌘⌥. to describe what you want in plain English and get a ready-to-run shell command
• Next-command suggestions — 3 smart suggestions after every run, shown as clickable chips
• No API key, no subscription, no cloud — AI runs entirely on your Mac via llama.cpp
• Choose from 7 bundled models: Qwen3 4B/8B, Gemma 3 4B, Qwen2.5-Coder 3B, and more
• Download, switch, and delete models from Preferences → AI

• /ls — file explorer with preview
• /ports — live network port inspector
• /run — Makefile & npm scripts picker
• Define your own custom command in Preferences with dynamic item lists — see examples →

• Multiple tabs with per-tab accent colors
• Split panes horizontally or vertically (⌘D / ⌘⇧D)
• Smart prompt segments — git branch, Node, Go, Python, Docker, K8s status
• Click any prompt segment to copy its value
• Smart command history with prefix filtering
• WebGL-accelerated rendering via xterm.js

• Check for updates — version shown in status bar, update notification with one-click upgrade instructions
• Long command support — commands of any length sent via temp file to bypass macOS PTY limits
• 7 built-in themes: Catppuccin Mocha, Tokyo Night, Dracula, Nord, Gruvbox, One Dark, Monokai

brew tap jebi-sh/tap
brew install --cask jebi

Or download the latest DMG from the releases page.

Requirements: Apple Silicon (M1 or later) · macOS 14 Sonoma or later

First launch: macOS may show a warning that jebi is from an unidentified developer. To open it, go to System Settings → Privacy & Security and click Open Anyway. This is expected for apps not distributed through the Mac App Store.

Prerequisites: Node.js 18+, Go 1.21+, Make

git clone https://github.com/jebi-sh/jebi.git
cd jebi
make deps   # download llama.cpp binaries (~30s)
make dev    # run in development mode
make build  # package the app (DMG + ZIP)

jebi ships with a model registry you can manage from Preferences → AI:

Models are downloaded on demand and stored in ~/Library/Application Support/jebi/models/.

• Custom Slash Commands — schema reference, command types, and examples for Git, Docker, Node, and more