Open AI3 mins read

OpenAI rolls out GPT-5.6 with ChatGPT Work, an agent for full workflows

OpenAI is launching ChatGPT Work, a Codex-powered agent inside ChatGPT that runs on GPT-5.6 and can work across business apps, files, and longer projects.

ChatGPT Work turns ChatGPT into a workflow agent

OpenAI is launching ChatGPT Work alongside the public rollout of GPT-5.6. The product is described as an agent inside ChatGPT that combines Codex technology with third-party integrations and can work across apps and files for hours on complex projects.

The core promise is finished work from a single prompt, such as producing an Excel or Word document from a task. For users, the practical takeaway is that OpenAI is positioning ChatGPT less as a chatbot and more as an autonomous work system.

Unified Plugins Directory brings business apps into one place

ChatGPT Work includes a new Unified Plugins Directory for third-party integrations. At launch, the directory includes services such as Google Drive, SharePoint, Slack, Microsoft Teams, Gmail, Outlook, Salesforce, Adobe, Zoom, LinkedIn, GitHub, Canva, and Dropbox.

Users can call specific plugins with an “@” mention or let ChatGPT determine which data source is relevant. That makes data access and app selection a central part of the workflow experience.

Rollout depends on platform and subscription tier

On web and mobile, Pro, Enterprise, and Edu users get ChatGPT Work first, with Plus and Business customers set to follow in the coming days. Through the ChatGPT desktop app for Mac and Windows, the product is available right away for all plans, including the free tier.

The article notes that, as of writing, the Windows link on the download page still leads to the Codex app. Readers should check their platform and plan before assuming access is available in the same way everywhere.

Billing and safety are built around agent-style usage

ChatGPT Work uses the same agent-based consumption pool as Codex, ChatGPT for Excel, and Workspace Agents. Usage depends on task size, complexity, and the model selected, while Enterprise and Edu admins can manage limits with Spend Controls at workspace, group, and individual levels.

OpenAI also points to an Auto-Review feature where advanced models check important actions before they run. According to the article, OpenAI says Auto-Review blocked 100 percent of attempts to extract protected data during adversarial red-teaming.

Discover More