LLM agents are unpredictable.
Their coordination
doesn't have to be.
omar delivers deterministic, formally specified orchestration
for multi-agent systems.
Features
Deep hierarchies
Agents managing agents, just like a company. Create parallel organizations of arbitrary depth.
Heterogeneous backends
Let Claude, Codex, Cursor, Opencode, and Google Antigravity collaborate as a team. Mix and match the best model for each task.
Full control
Talk to and control any subagent you want. Navigate hierarchies with arrow keys and attach via popup.
Flexible lifespans
Long-running or ephemeral agents, your choice. Persistent state across sessions with memory snapshots.
Messaging & integrations
Connect agents to Slack channels. Bridge the gap between human and AI workflows seamlessly.
Built on tmux
Supports all tmux commands and key bindings. Highly customizable with battle-tested session management.
Quick Start
$ curl -fsSL https://omar.rs/install.sh | sh
Works on macOS and Linux. Installs all binaries to /usr/local/bin.
$ brew install omar-os/omar/omar Automatically installs tmux as a dependency.
$ git clone https://github.com/omar-os/omar.git
$ cd omar && make install Requires Rust 1.70+ and GNU Make.
Prerequisites
tmux 3.0+ and at least one agent backend: Claude Code, Codex CLI, Opencode, Cursor CLI, or Google Antigravity CLI.
Launch Mission Control
$ omar serve --ui Opens the web UI in your browser. Describe a workflow, confirm the program it drafts, and watch the run on a live diagram.
Or drive it from the terminal
$ omar The same runtime in tmux. Copy a prompt into your Executive Assistant window and watch agents spawn.