OpenAI introduced ChatGPT Agent on July 17, 2025. The announcement says access began rolling out that day, but not to every plan at once: Pro users were expected to get access by the end of the day, Plus and Team users over the next few days, and Enterprise and Education users in the coming weeks. The announcement does not establish country-by-country or account-by-account access timing, so July 17 should not be read as proof that every user could access it immediately. (OpenAI announcement)
OpenAI’s current Help Center says the original ChatGPT Agent feature is no longer available and directs users to ChatGPT Work for longer, multi-step tasks and finished deliverables. (OpenAI Help Center)
Release date and initial availability
OpenAI’s announcement is dated July 17, 2025, and describes the feature as starting to roll out that day. Its availability guidance specifies the rollout by plan:
| Event | Timing | What it establishes |
|---|---|---|
| ChatGPT Agent introduced; rollout begins | July 17, 2025 | OpenAI’s announcement date and the start of rollout—not immediate access for every account or region. |
| Pro access | By the end of July 17, 2025 | The stated timing for Pro users. |
| Plus and Team access | Over the next few days | A staggered rollout after the initial announcement. |
| Enterprise and Education access | In the coming weeks | A later rollout for these plans. |
The July 17 date is the right answer for when ChatGPT Agent was introduced. The rollout was staged by plan; the announcement’s reported timing does not establish region-specific or account-by-account access. (OpenAI announcement)
Is ChatGPT Agent available through a production API?
OpenAI’s launch and Help Center material do not identify a production API for invoking the original ChatGPT Agent feature. The launch describes agent mode inside ChatGPT; that is different from a developer API for building an agentic application. (OpenAI announcement)
For a developer building an application, OpenAI’s Responses API is the documented route to evaluate. Requests use the POST /v1/responses endpoint and OpenAI API authentication; OpenAI recommends Responses for new projects. This lets developers build their own workflows with supported models and tools—it does not invoke the original ChatGPT Agent feature. (Responses API migration guide; production best practices)
OpenAI has also introduced a separate Agents API in public beta for building applications with an OpenAI-managed Codex agent. It is a distinct product, not an API for the original ChatGPT Agent feature; because it is labeled beta, assess its current suitability and terms before relying on it for production. (OpenAI announcement)
Practical choice: Use the Responses API as the documented starting point for a new agent application, and check its current API documentation and access requirements for your project. Don’t assume that a ChatGPT subscription feature and a developer API provide interchangeable access.
Sources
Introducing ChatGPT agent | OpenAI
ChatGPT agent | OpenAI Help Center
Migrate to the Responses API | OpenAI API