OpenAI is expanding ChatGPT's capabilities with the launch of "workspace agents." This new feature allows businesses to automate recurring tasks, even when employees are offline.
Powered by OpenAI's Codex model, these agents act as persistent assistants. They can connect to external applications, retain project information, and execute multi-step workflows without continuous prompting. OpenAI states these agents can handle tasks like report preparation, code writing, and message responses.
Workspace agents run in the cloud, enabling continuous operation. They are designed for organizational sharing, allowing teams to build, use collaboratively in ChatGPT or Slack, and refine agents over time.
Users can create an agent by describing a desired workflow within ChatGPT, which then assists in mapping processes, connecting tools, and testing the agent. Once active, agents can operate on schedules or respond to specific triggers.
This development arrives amidst a significant investment phase in agentic AI, with major tech companies pouring billions into autonomous systems. OpenAI assures that companies can limit agent data access, require human approval for sensitive actions, and monitor for security threats like prompt injection.
The workspace agents are currently in a free research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. This preview period ends May 6, 2026, after which a credit-based pricing model will be implemented.
OpenAI's own teams are already utilizing this technology. The company also confirmed that existing GPTs will remain available and can be easily converted into workspace agents.