omg.dev

What problem omg.dev solves

omg.dev is an open source agent control plane. The README says you run your AI coding agents on your own machine and drive them from your phone. The stated problem is practical. Running one coding agent in a terminal is fine, but running five is not, because they die when you close the laptop, you cannot tell which one is stuck waiting on a permission prompt, and you have to be at your desk to answer it. omg.dev turns a Linux box or a macOS workstation into a private control plane for several agents. The README names Claude Code, Codex, OpenCode, Jcode, Cursor, Grok, fx, Pi, and GitHub as the agents it can control. The project is described as the open source agent control plane, and you can self host it or use the hosted service at omg.dev. The MIT license badge in the README means the code is open for self hosting and modification. The core benefit is that your agents keep running on a machine that stays up, and you can check on them and answer prompts from a phone instead of being tied to a desk. This addresses the gap between a single local terminal session and a fleet of agents that need oversight when you are away from the keyboard.

How the remote access works

The README organizes its documentation around the mobile use case. There is a quick start section, a section on letting an agent install it, a Discord link, a why omg.dev section, an agents section for connecting a coding agent, a remote access section for reaching it from your phone, and a security section. The introduction states you drive the agents from your phone, which implies a client that talks to the control plane running on your own machine or on the hosted service. Because the control plane runs on a Linux box or macOS workstation, the agents are not bound to your laptop being open. The phone client is what lets you see which agent is stuck and answer its permission prompt while you are away. The README's screenshot set shows a web UI, a scheduled agents view, and a usage limits view, which suggests the interface surfaces status and quotas, not just a raw terminal. The security section is where the project explains how access is protected, which matters because you are exposing agent control beyond your local terminal. For a user, the flow is to install omg.dev on an always on machine, connect the coding agents you use, and then open the mobile client to monitor and steer them from anywhere with a connection.

How to get started and stay safe

The README points to a quick start as the entry point for installing omg.dev, and it also offers a path where you let an agent install it for you. The project is MIT licensed, so you can self host the control plane on your own infrastructure instead of using the hosted service. The agents section explains how to connect a coding agent, and the list of supported agents includes Claude Code, Codex, OpenCode, Jcode, Cursor, Grok, fx, Pi, and GitHub. The remote access section covers reaching the plane from your phone, and the security section is the place to read before exposing the service. Because the tool can answer permission prompts on your behalf, the security model is important, and the README flags it as a dedicated topic rather than an afterthought. The hosted option at omg.dev is for users who do not want to run their own server, while self hosting keeps the agents on your machines. The Discord link gives a community channel for questions. For a first run, the practical steps are to read the quick start, choose self host or hosted, connect at least one coding agent, and then install the phone client so you can manage sessions away from your desk while reviewing the security notes first.

Editorial conclusion

omg.dev is written in TypeScript under the MIT license, and you can either self host it or use the hosted service at omg.dev to manage agents such as Claude Code, Codex, OpenCode, Cursor, and Grok.