I wanted to work with my agent sessions in a regular browser and use the whole screen to follow the work. When I started, I could control my Codex sessions remotely through the ChatGPT app on a phone or tablet, but I missed that option on chatgpt.com.
I wanted an overview of Git changes, what was happening in each session, how much of the context window was in use, and token usage. And I wanted to start new sessions directly from the browser, without opening more terminals first.
It started as a browser interface for Codex. With remote-agents, it has become a shared workspace for both Codex and Claude, with persistent terminals alongside the agent sessions. I can start work, return to it and follow its progress on a full screen.
remote-agents brings further development together in one project. That makes more sense to me than maintaining two solutions for the same need.