Getting Started
What is OMAR?
OMAR is a TUI for creating powerful agentic organizations. It runs on tmux and lets you orchestrate deep hierarchies of parallel AI agents from a single terminal.
Prerequisites
- tmux 3.0+ —
brew install tmux(macOS) orapt install tmux(Debian/Ubuntu) - At least one agent backend: Claude Code, Codex CLI, Opencode, Cursor CLI, or Gemini CLI
Installation
One-Liner (Recommended)
$ curl -fsSL https://omar.tech/install.sh | sh
Works on macOS and Linux. Installs all binaries to /usr/local/bin.
Homebrew
$ brew install lsk567/omar/omar Automatically installs tmux as a dependency.
From Source
$ git clone https://github.com/lsk567/omar.git
$ cd omar && make install Requires Rust 1.70+ and GNU Make.
Launch
$ omar This starts the OMAR dashboard with the default agent backend (Claude Code). To use a different backend:
| Backend | Command |
|---|---|
| Claude Code | omar or omar --agent claude |
| Codex CLI | omar --agent codex |
| Opencode | omar --agent opencode |
| Cursor CLI | omar --agent cursor |
| Gemini CLI | omar --agent gemini |
Your First Session
-
Launch
omar— you'll see the TUI dashboard with your Executive Assistant (EA) -
Copy a test prompt into your EA window:
Load and run https://github.com/lsk567/omar/blob/main/prompts/tests/project-factory.md - Watch agents spawn. Use ↑↓ to navigate, Tab to drill into an agent, Shift+Tab to back out
-
To shut down:
Shutdown the test project and its agents.
Key Bindings
| Key | Action |
|---|---|
←/→ or h/l | Switch panel (sidebar / main) |
↑/↓ or j/k | Move selection up/down |
Tab | Drill into selected agent |
Shift+Tab or Esc | Back (drill up) |
Enter | Attach to selected agent |
n | Spawn new agent |
d | Kill selected agent |
N | Spawn new EA (prompts for name) |
D | Delete current EA |
[ | Previous EA |
] | Next EA |
p | Add a project |
e | Show scheduled events |
S | Settings |
G | Debug console |
r | Refresh agent list |
z | Detach (dashboard keeps running) |
? | Toggle help |
Q | Quit |
Next Steps
- Architecture & Design — understand how OMAR works
- HTTP API Reference — programmatic agent control
- Agent Orchestration — manager-worker patterns