ChatGPT gains agent mode

OpenAI has added a formal agent mode to ChatGPT, turning the chatbot into a tool that can browse, click, fill forms, edit files, and take other actions for users. (help.openai.com) OpenAI’s help page says ChatGPT agent can navigate websites, work with uploaded files, connect to third-party data sources such as email and document repositories, and edit spreadsheets while the user stays in control. The feature is selected from ChatGPT’s tools menu. (help.openai.com) A day later, on April 22, 2026, OpenAI introduced workspace agents for ChatGPT Business and Enterprise. OpenAI said the agents are powered by Codex, can run long workflows in the cloud, and can be shared across a company in ChatGPT or Slack. (openai.com) OpenAI described workspace agents as an evolution of Custom GPTs: instead of answering prompts inside one chat, they can prepare reports, write code, respond to messages, and keep working after the user closes the window. Enterprise and Business release notes said the rollout would take place over the following weeks. (openai.com (help.openai.com) The change extends a broader shift already underway inside ChatGPT. OpenAI renamed “connectors” to “apps” in December 2025 and said those integrations let ChatGPT search, reference, and work with information from outside services without leaving the conversation. (help.openai.com) OpenAI’s release notes show that shift accelerating in 2026. On March 27, the company said it updated Box, Notion, Linear, and Dropbox apps with new actions, including write capabilities where supported, and Business release notes later added synced Notion and Linear connectors for team chat. (help.openai.com 1) (help.openai.com 2) That means ChatGPT is moving from a place to ask questions into a place where company data, software tools, and automated tasks meet. OpenAI’s business page says workspace agents can operate across Slack, Google Drive, Microsoft SharePoint, and other enterprise tools with admin controls. (openai.com) OpenAI has paired that expansion with more administrative controls. Help Center pages referenced in OpenAI search results say Enterprise admins can enable or disable agent mode, and OpenAI has published separate documentation for allowlisting agent traffic to company sites and networks. (help.openai.com) OpenAI is also tying the rollout to Codex, its product for long-running coding and multi-agent work. OpenAI’s Codex pages describe a system built for parallel tasks, cloud environments, reviews, refactors, and releases, which helps explain why the company is now using the same engine for workplace automation inside ChatGPT. (openai.com 1) (openai.com 2) The immediate result is that ChatGPT now spans three layers at once: chat, connected apps, and agents that can act. OpenAI’s own product pages frame the next step as shared automation inside company guardrails, not just better answers in a text box. (help.openai.com) (openai.com)