OpenAI's Workspace Agents Turn ChatGPT into a Shared Team Worker

OpenAI killed the GPT-as-chatbot framing today. Workspace Agents in ChatGPT, launched April 22, are Codex-powered agents that teams build once, share, and improve together. They run in the cloud, keep working when you close the tab, take real actions across connected tools like Slack, and they live inside an organization's permissions and audit controls.

The research preview is free until May 6, then moves to credit-based pricing. Available in Business, Enterprise, Edu, and Teachers plans. Admins get to scope which connected tools each user group can touch and who has rights to build, share, or use a given agent.

The positioning is the part to read carefully. Custom GPTs were a personal toy. Workspace Agents are a shared team member with a job description, a permission set, and continuous improvement loops. This is the OpenAI answer to Google's Agent Studio and Anthropic's Claude Cowork pushed into the same week as Cloud Next. The agentic work surface is now a contested category, and the pricing model is starting to look like SaaS seats with metered actions on top. Source: https://openai.com/index/introducing-workspace-agents-in-chatgpt/