Running Claude Code and Codex from your phone: the official options, and the self-hosted app I built A developer built TaroBird, an open-source (AGPL-3.0) self-hosted chat app that runs coding-agent CLIs — Codex, Claude Code, Kimi Code, DeepSeek Harness and Antigravity — on the user's own machine and lets them be driven and handed off between each other from iOS and Android phones. The app does not reimplement the agents or proxy requests, running each official CLI under the user's own sign-in and subscription. The post also reviews the vendors' official mobile options: Claude Code's `/remote-control` mode and Codex in the ChatGPT mobile app, both of which stay within a single vendor. Disclosure: I'm the author of TaroBird, the open-source app in the second half of this post. The official options come first because for many people they are enough. If you run Claude Code or Codex for long tasks, you know the pattern. You start something, walk away, and the agent stops ten minutes later to ask for a confirmation. By the time you are back at your desk, the run has been idle for an hour. There are now several ways to keep those runs moving from a phone. Here is what I learned using them, starting with what the vendors ship themselves. Claude Code has a built-in Remote Control mode. Inside a running session you type /remote-control or /rc , or you start a new session with claude --remote-control . The session keeps running on your computer, and you can continue it from claude.ai/code or the Claude mobile app. Things worth knowing, from the official docs https://code.claude.com/docs/en/remote-control : /config . If Claude Code is the only agent you use, this is the first thing to try. It needs no extra software. OpenAI added Codex to the ChatGPT mobile app. You pair your phone with the ChatGPT desktop app on your Mac or Windows PC by scanning a QR code, and from then on you can start tasks, approve actions and review results from the phone while the work runs on your computer. The setup guide https://learn.chatgpt.com/docs/mobile also reminds you to keep the computer awake and online. Again: if Codex is your main agent, start here. My own setup is messier. On a normal day I have Codex, Claude Code, Gemini through the Antigravity CLI , Kimi Code and DeepSeek Harness running on the same Mac, and sometimes on a second machine. Each official option covers its own vendor, so I ended up juggling apps, and none of them let one agent hand work to another. So I built TaroBird. TaroBird is an open-source AGPL-3.0 , self-hosted chat app for coding agents. Each "Bot" is one of the official CLIs running on your own computer, under your own sign-in and subscription. TaroBird does not reimplement the agents, extract tokens or proxy requests. Your skills and configuration are the same as in your terminal. What that gives you in practice: | | Claude Code Remote Control | Codex in the ChatGPT app | TaroBird | |---|---|---|---| | Agents | Claude Code | Codex | Codex, Claude Code, Kimi Code, DeepSeek Harness, Antigravity | | Where work runs | Your computer | Cloud or your computer | Your computer or your own server | | Phone app | Claude app | ChatGPT app | TaroBird app iOS, Android | | Handing work between different vendors' agents | Not across vendors | Not across vendors | Yes | | Extra software | Claude app on the phone | ChatGPT desktop and mobile apps | TaroBird desktop and mobile apps | | Source | Closed | Closed | Open source AGPL-3.0 | The code, installers and docs are on GitHub: https://github.com/cloveric/tarobird https://github.com/cloveric/tarobird If one vendor's official remote option already covers how you work, use it. If you run several agents and want them in one place, and talking to each other, TaroBird might be worth a try. I'd like to hear what breaks.