How to Use Codex from Your Phone (2026)

You kick off a Codex run on your Mac before lunch. Something big — refactor the API layer, touch a few dozen files. You close the laptop lid, or just walk to another room, and come back forty minutes later. Codex isn’t done. It stalled at minute four on a permission prompt — “allow this command?” — and has been sitting there ever since. Or worse: it kept going, confidently, down a path you’d have killed in five seconds if you’d been watching.

This is the actual problem with long-running agents. Not that they can’t work — that they can’t work unsupervised, and you can’t always be at the keyboard. If you could watch and steer Codex from the phone already in your pocket, none of those forty minutes would have been wasted.

So: what are your options for running Codex from your phone in 2026? Here’s an honest survey.

The ways to run Codex from your phone

There are broadly three approaches, and they’re aimed at different people.

1. The official Codex mobile experience (inside the ChatGPT app). OpenAI shipped mobile access to Codex in 2026. Codex itself runs on a machine you connect — your laptop, a Mac mini, or a remote environment you’ve set up — and the phone becomes a control surface: start a task, steer it while it runs, approve actions, and check in on progress. It’s the most first-party option, and if you already live inside ChatGPT it’s the path of least resistance.

2. Third-party direct-connect apps. A small category of tools connect your phone directly to a Codex instance running on your own machine, without routing your work through a hosted product. These vary a lot in maturity and in how much of the agent’s activity they actually surface.

3. SeaWork. That’s us. We built SeaWork to connect your phone (and desktop, and terminal) directly to the Claude Code and Codex agents already running on your local machine — with the code staying on your machine. More on where that’s different below.

None of these is strictly “best.” They’re trade-offs between convenience, where your code lives, and how many agents and providers you want to manage. Pick based on what you actually care about.

What the official Codex mobile does — and doesn’t

Credit where it’s due: the official experience is genuinely useful. You can start a Codex task from your phone, steer it mid-run, approve the actions it wants to take, and see how it’s progressing without being at your desk. For a lot of people that’s exactly enough.

The trade-offs are mostly about product surface and account model. Official Codex mobile is part of ChatGPT and pairs with a signed-in Codex host; it’s excellent if your workflow is already Codex-first. What it doesn’t do is manage Claude Code or other local agent CLIs from the same mobile surface.

If those constraints fit you, it’s a fine answer. If they don’t, keep reading.

When you want the agent on your machine, not a hosted environment

Here’s the first-principles argument for keeping the agent local, stripped of any pitch.

Codex on your laptop is already wired into your world. It has your credentials. It talks to your MCP servers. It reads and writes your actual files, in your actual repo, with your actual environment variables and tool versions. That setup took you time to build, and it’s correct.

Some hosted-agent workflows do require recreating your setup elsewhere — syncing credentials, rebuilding MCP and tool config, giving a remote environment access to your repo. If your priority is keeping the agent process on the machine where your environment already works, the cleaner pattern is to move the interface, not the agent.

So the harder-but-honest path is: don’t move the agent. Move the interface. That’s the principle SeaWork is built on — the agent stays where the work already is; the control surface comes to you.

Concretely: the SeaWork daemon runs on your development machine. Your phone connects to it over an end-to-end encrypted relay. The relay only forwards encrypted packets — it can’t read your prompts, your diffs, or your logs, because it never has the keys. It can still see connection metadata — IPs, timing, message sizes, session IDs — but never the contents. Your code, logs, and provider credentials stay on your machine. If you’d rather not depend on any relay at all, SeaWork is self-hosted by nature: the daemon is yours, and you can run the whole thing on your own network. Self-hosted Codex, controlled from your phone, without the cloud middle layer.

One app for Codex and Claude Code

The official Codex mobile path is Codex-first — which is fine until your workflow also leans on Claude Code.

In practice, plenty of us don’t pick one model and stay there. You reach for Codex for some tasks and Claude Code for others, sometimes in the same afternoon, sometimes on the same feature. SeaWork runs both from a single app — Claude Code and Codex, side by side, on your phone or desktop.

It also runs them in parallel. One agent refactoring the API layer, another writing the tests for it, a third drafting the migration notes — each in its own workspace, each visible at a glance so you can see what’s running, what’s waiting on you, and what finished. For anyone splitting implementation, tests, and review across separate worktrees, a single terminal scrollback stops scaling fast. A phone that shows you every agent at once fits that workflow far better than a single scrollback buffer does.

Set it up in three steps

You don’t need to install a separate SeaWork daemon — the desktop app ships with it. (SeaWork drives the agents you already have, so Codex or Claude Code should already be set up and working on that machine.)

1. Install the desktop app. It ships with the SeaWork daemon built in, so your dev machine is ready with nothing else to set up.
2. Pair your phone. Open the mobile app and scan the pairing code. That establishes the end-to-end encrypted link between phone and daemon.
3. Start a Codex agent. From the phone, the desktop, or the terminal.

The CLI is the fastest way to show what’s happening under the hood. These are real commands:

# Start a Codex agent on a task
seawork run --provider codex "refactor the API layer"

# List your agents (running and finished)
seawork ls -a

# Follow an agent's logs live
seawork logs <id> -f

Start something like that, then pick it up on your phone. You’ll see every tool call Codex makes as it makes them, read the full conversation, and send a message mid-run to course-correct when it drifts. And when Codex hits an approval prompt — the “allow this command?” gate that would otherwise freeze the run until you’re back at your desk — the request lands on your phone and you approve or reject it in seconds. That’s the forty wasted minutes from the top of this post, gone.

Where to start

Running Codex from your phone in 2026 is no longer exotic — the honest question is which version of it fits how you work. Check Codex pricing and plan limits first, then compare Codex CLI vs Claude Code if your workflow spans both agents. If you want the agent to keep running where your code already lives, controlled from your pocket, across both Codex and Claude Code, that’s the corner we built for.

SeaWork is free. Download SeaWork, read more about Codex on SeaWork, or follow the getting started guide to run your first agent.