OpenAI workspace agents

OpenAI has started rolling out workspace agents in ChatGPT, a new shared bot for teams that can keep working across apps after a chat ends. (openai.com) OpenAI announced the feature on April 22, 2026, and said it is powered by Codex, the company’s coding and task-running system. The agents run in the cloud, can handle long workflows, and operate inside the permissions set by a company’s administrators. (openai.com) The rollout is gradual over the next few weeks for ChatGPT Business and Enterprise workspaces, according to OpenAI’s release notes. OpenAI’s help center says the same system also supports Slack, where teams can create, use, and manage agents outside the main ChatGPT app. (help.openai.com 1) (help.openai.com 2) A workspace agent is meant for repeatable office work, not a one-off prompt. OpenAI says teams can build one from scratch or from templates, connect it to tools like Google Drive, Google Calendar, Slack, and SharePoint, then let it run on a schedule. (openai.com) (help.openai.com) That changes the role ChatGPT has played inside companies over the past year. OpenAI says earlier GPTs helped individuals move faster, while workspace agents are built for jobs that depend on shared context, handoffs, approvals, and actions across multiple systems. (openai.com) OpenAI is not shutting down custom GPTs. The company said workspace agents are “an evolution of GPTs,” and reports from The Verge and other outlets say OpenAI plans a way to convert existing GPTs into persistent team agents later. (openai.com) (theverge.com) OpenAI’s business page says the agents can automate multi-step workflows across Slack, Google Drive, Microsoft SharePoint, and other enterprise tools with governance controls. VentureBeat reported that OpenAI is also positioning Salesforce as one of the systems these agents can use for customer-record data and actions. (openai.com) (venturebeat.com) OpenAI is packaging the feature as part of a broader push to make ChatGPT a workplace platform, not just a chatbot. Its product release log shows workspace agents landing one week after an update to the Agents software development kit and six days after “Codex for (almost) everything.” (openai.com) The immediate test is whether companies trust a shared agent to pull data, ask for approvals, and act inside work software without creating new security headaches. OpenAI’s pitch is that this bot lives inside the same admin controls and workspace rules that already govern the rest of a team’s ChatGPT use. (openai.com) (help.openai.com)