OpenAI Launches Workspace Agents: Shared Codex-Powered Bots for Teams

OpenAI launched Workspace Agents in ChatGPT on April 22. The feature is a successor to Custom GPTs, but designed for team use and longer-running tasks rather than individual customization.

Workspace Agents are powered by Codex and can handle tasks like writing code, preparing reports, and responding to messages. The key difference from Custom GPTs is that Workspace Agents run in the cloud continuously, meaning they can keep working on a task after you close ChatGPT and pick up where they left off.

They’re also built for sharing. Teams can build an agent once and deploy it across the organization through ChatGPT or Slack. When the team identifies improvements, updates apply across everyone using it.

What They Can Do

The initial feature set covers:

• Running multi-step workflows that span long time periods
• Continuing work autonomously while the user is offline
• Integrating with external services like Slack and Salesforce
• Operating within organization-defined permissions and controls

The Codex connection matters here. Codex’s architecture is designed for agentic tasks, running headless operations, managing context across large codebases, and executing code in loops. That infrastructure is what gives Workspace Agents the ability to stay running independently rather than requiring an active session.

OpenAI is also committed to a conversion path for existing Custom GPTs, so organizations that have already built GPT customizations can upgrade them to Workspace Agents.

Availability and Pricing

Workspace Agents are in research preview now, available in ChatGPT Business, Enterprise, Edu, and Teachers plans. Free and Plus subscribers don’t have access yet.

The feature is free until May 6. After that date, credit-based pricing kicks in, though specific rates haven’t been announced.

OpenAI has outlined upcoming additions: triggers to start workflows automatically without manual invocation, dashboards to track and optimize agent performance, more third-party integrations, and support for Workspace Agents in the Codex desktop app.

Who This Is For

Workspace Agents sit between individual productivity tools and full enterprise automation. A developer team could build an agent that runs code review on incoming PRs, an operations team could build one that monitors metrics and drafts daily reports, or a support team could build one that handles routine ticket classification.

The Codex connection makes it particularly relevant for software engineering workflows. Codex crossed 4 million weekly active developers earlier this month, and OpenAI has been building out enterprise distribution through GSI partnerships to scale adoption. Workspace Agents extend that infrastructure to team-level automation.

The research preview label means the feature is functional but not finalized. Organizations that want to evaluate it before the paid tier kicks in on May 6 have a window to do so now.

---

Sources: 9to5Mac, OpenAI Academy, AIToolly