Overview
OmO is a multi-model agent orchestration harness designed to turn one AI session into a coordinated team for practical work. It supports software development, research, reports, presentations, experiments, and other tasks that benefit from planning, delegation, and verification.
The main agent remains responsible for the overall job while routing specialized work to category workers or focused agents. OmO can use models from different providers rather than requiring a single vendor.
Key Features
- Routes tasks to specialized workers for areas such as frontend development, debugging, research, data science, visual quality, and code review.
- Provides planning and execution modes, including interview-based planning followed by work in the same session.
- Uses tuned model profiles that select an available main-session model according to the chosen intent.
- Supports independent review and memory recall through a read-only sidecar that can warn the main agent about relevant prior context.
- Verifies work with diagnostics and checks before considering a task complete.
How It Works
A user describes a task to the main agent, which plans the work and delegates parts through task categories or dedicated exploration and documentation agents. Each delegated worker receives a fresh session with the model and skills configured for that category, while the main agent integrates and verifies the results.
Profiles can choose a general-purpose or deep-reasoning model from connected providers. An exact provider and model can also be pinned, while explicit command-line or scoped model choices take precedence.
Installation and Use
OmO provides a native command-line binary for macOS, Linux, WSL, and Windows, without requiring Node.js, npm, or Bun. The installer selects the correct build for the operating system, processor, and C library, verifies it against release checksums, and installs it in the user's local binary directory. After installation, users open a project, run the OmO command, and describe the work to perform.