A new tool syncs OpenAI Codex sessions between a desktop and a laptop

OpenAI Codex doesn't have built-in sync between devices yet, so moving from a home desktop to a laptop can wipe out your working momentum. A community-built tool called codexSync offers a practical workaround: instead of relying on a "Codex cloud," it moves Codex's local state through a separate sync folder.

Why this hurts

The problem shows up quickly if you use Codex in two common setups: at your main computer and on the road. Over time, the assistant accumulates conversations, intermediate states, notes, summaries, and project context. All of it stays tied to one specific device.

As a result, a task you start on one computer doesn't pick up seamlessly on another. You can't simply pull in the history, merge it, or clean it up properly — the only option is a crude archive.

Started a task on the desktop → can't continue it on the laptop.

Based on the author's observations, the cause is that Codex stores its working state locally — in session files, internal data, and possibly SQLite. That design has obvious upsides: speed, offline operation, and privacy. The downside is the lack of a proper multi-device workflow. On a single machine you barely notice it, but for people who constantly switch between a desk and travel, it becomes a real productivity limit.

How the workaround works

The solution doesn't try to hack Codex itself or hook into its internal logic. Instead, it syncs the state folder between devices and works on top of the existing storage model.

To do this, the author built a small CLI utility, codexSync. It touches the data no more than necessary and uses a shared directory, which can be cloud storage, a NAS, or any other accessible folder.

• automatically finds the relevant Codex directories
• syncs sessions and related internal files
• supports config validation, a change plan, and dry-run
• lets you manually pull state before you start working and push changes afterward

Getting started is simple. First, you create a config with paths to the .codex directory and the sync working folder. The utility then validates the settings, shows a plan, and, if needed, runs a trial pass without writing anything.

The daily workflow is just as straightforward: before a session, you pull the latest state, and when you're done, you push your changes back. This mode doesn't deliver real-time magic, but it solves the basic portability problem.

Where the limits are

Crucially, this is not an official OpenAI feature and not full real-time sync. To avoid corrupting the state, you have to fully close Codex before syncing, including background processes.

If the app is still running, the sync shouldn't start; the author even added a mechanism that offers to kill a hung process. That makes the tool more of a disciplined work routine than an invisible background service.

At first glance, you might ask why you need a dedicated utility at all when you could just drop the folder into Dropbox, OneDrive, or another cloud drive. But that approach has too many weak spots: file conflicts, unpredictable states, no transparent change plan, and the risk of corrupting the local database. codexSync adds a layer of control where a plain cloud folder offers only blunt syncing with no idea of what exactly is about to be overwritten.

The tool still needs testing on different configurations, especially macOS ↔ macOS and macOS ↔ Windows setups. That matters: the idea itself looks universal, but path handling, file locking, and background processes can behave differently from one system to another.

So for now, the project looks more like a useful open-source tool for power users than something you can recommend to every Codex user without caveats.

What it means

The codexSync story illustrates a simple fact: a layer of user-built infrastructure quickly grows around AI tools to fill gaps in official products. If OpenAI doesn't add native sync to Codex, external utilities like this will become the standard way to keep work continuous across devices.