ChatGPT Work and Codex to auto-sync agent work

What happened

On August 11, OpenAI's developer account unveiled a feature that syncs work between ChatGPT Work and Codex. Users can import projects, conversations, skills, and plugins built up in one agent into the other, and check import history from settings. Auto-updates are opt-in, so users must turn them on manually for them to take effect. The feature is currently available only on the ChatGPT desktop app.

What this means

Codex is OpenAI's coding agent, and ChatGPT Work is its enterprise ChatGPT plan. The two products previously kept separate work histories and settings. But the broader agent ecosystem has recently been settling into a pattern where a single person uses multiple tools at once. On August 9, Anthropic added budget caps and an advisor model feature to Claude's managed agents, and on August 10, Microsoft released guidance for building Copilot agents without code. Building skills and plugins and accumulating conversation context tends to repeat across tools, and redoing this work each time is inefficient. OpenAI's new sync feature reads as an attempt to eliminate this duplication, at least within its own ecosystem. This also lines up with OpenAI's push in August to make agent features a pillar of its paid offerings, having introduced a $125-per-month premium seat for ChatGPT Business.

What changes as a result

In practice, users will no longer need to manually rebuild skills or plugins created in Codex when moving them to ChatGPT Work. With auto-updates turned on, changes made on one side carry over to the other, easing management burden for companies running agents across teams. This extends OpenAI's broader move to expand ChatGPT from a single tool into a work platform where multiple agents operate in tandem.