OpenAI enables workspace agent billing
- OpenAI began charging for ChatGPT Workspace Agent runs on July 6, replacing the free preview with token-based pricing tied to workspace credit pools. - OpenAI’s help documentation says Enterprise and Edu admins can set monthly user limits, monitor shared credits, and cap or block overages. - OpenAI’s rate card and release notes now list Workspace Agent pricing, with Enterprise, Edu, and Business billing controls in admin settings.
OpenAI started charging for ChatGPT Workspace Agent runs on July 6, according to the company’s release notes and help documentation. The change moves the feature out of research preview and makes agent usage billable through token-based credits rather than bundled as a free trial feature. OpenAI also published new controls for ChatGPT Enterprise and Edu administrators to set monthly usage limits, monitor shared credits and manage overages. The update makes cost tracking part of day-to-day operations for companies using agents for internal workflows. ### When did OpenAI start charging for workspace agents? July 6, 2026 is the date OpenAI gives for the start of token-based pricing for Workspace Agent runs. OpenAI’s release notes say credit use is based on input tokens, cached input tokens and output tokens, and that Business plans include some usage through the broader Codex agentic usage pool before flexible pricing applies. OpenAI introduced workspace agents earlier as a research preview for ChatGPT Business, Enterprise, Edu and Teachers plans. An OpenAI product page said teams could build agents for repeatable workflows across connected apps such as Slack, Google Drive and Microsoft SharePoint. ### Which customers get the new billing and controls? ChatGPT Enterprise and Edu administrators can now set monthly limits for users, control workspace overage usage and monitor shared credits, according to an OpenAI Help Center article. (openai.com) The same documentation says those controls are available to workspace owners and admins through billing settings. (openai.com) ChatGPT Business customers are also affected by the pricing change. OpenAI’s Business release notes say Workspace Agent runs now use token-based pricing and can draw from the workspace credit pool after included usage is exhausted. ### How are the charges calculated? OpenAI’s rate card says Workspace Agent pricing is based on credits per 1 million input tokens, cached input tokens and output tokens. (help.openai.com) The same rate card says Workspace Agent pricing is now in effect, while PowerPoint usage remains free for Business and Enterprise customers through August 6, 2026 before moving to the same token-based model. (help.openai.com) OpenAI’s release notes repeat that the billing model is token-based rather than per seat or per run. That means the cost of an agent workflow can vary with prompt length, tool use and the amount of generated output. ### What can admins actually control now? Enterprise and Edu workspace owners can configure usage alerts and hard overage limits in billing settings, according to OpenAI’s flexible pricing documentation. (help.openai.com) OpenAI says that if overage limits are set to zero, an organization will not be allowed to go into overages. OpenAI also says that once the shared credit pool is exhausted, advanced features are paused unless workspace owners enable overages or buy additional credits through their account team. (openai.com) If no credits are available, the feature is blocked and users can send an in-product request to an administrator. ### Why does this matter for teams using internal automations? (help.openai.com) Token-based billing makes usage metering more immediate for teams that run agents inside recruiting, sales, support or operations workflows. OpenAI’s documentation ties Workspace Agents to connected business apps and shared credit pools, which means administrators can now see usage as a billable workspace resource rather than an experimental feature. (help.openai.com) OpenAI’s published controls do not prescribe how companies should manage that spend. But the new billing structure means teams can now set ceilings, watch credit balances and decide whether to permit overages before broader rollouts of agent-based workflows. ### What happens next? August 6, 2026 is the next listed pricing milestone in OpenAI’s rate card, when ChatGPT for PowerPoint is scheduled to move to the same token-based model for Business and Enterprise customers. (openai.com) Workspace owners who use agents today can find the current billing rules, limits and overage settings in OpenAI’s rate card, release notes and Enterprise and Edu billing documentation. (help.openai.com 1) (help.openai.com 2)