ChatGPT Workspace Agents
OpenAI
Codex-powered shared agents for repeatable, governed team workflows in ChatGPT and Slack.
Overview
Freshness note: AI products, preview access, and credit rules change quickly. This profile is a point-in-time snapshot last verified on July 10, 2026.
ChatGPT Workspace Agents are OpenAI’s reusable shared-agent layer for teams. They run in the cloud, work across connected tools, and can be published across an organization so a repeatable workflow does not have to be rebuilt inside every conversation.
The July launch of ChatGPT Work makes the distinction important. Work is a personal agent for completing a substantial task and returning an artifact. Workspace Agents are organization-owned workflows with shared instructions, connections, schedules, Slack deployment, analytics, collaborator editing, and admin controls.
Key Features
Teams can create an agent from a recurring job, connect its tools and context, define approval points, test it, and share it through ChatGPT or Slack. OpenAI’s examples include software-request review, product-feedback routing, weekly metrics reporting, lead outreach, and third-party risk screening.
Workspace Agents can keep working in the cloud, respond to Slack requests, and run on a schedule. Supported API triggers can start a published agent from another system, although the trigger surface is not a general synchronous workflow API that returns the completed result inline.
Governance is more explicit than in an ad hoc chat. Admins can control who may use, build, or share agents; connected systems retain their own permissions; sensitive actions can require approval; and analytics expose usage patterns. Enterprise and Edu controls also extend into role-based access and supported Compliance API coverage.
Strengths
Workspace Agents are strongest when a team already understands the workflow and wants to make it reusable. Shared instructions, context, and approval rules reduce the drift that appears when every operator maintains a personal prompt version.
Slack deployment is also practical for workflows that should meet people where requests already arrive instead of forcing every user into a separate builder interface.
Limitations
The product remains a research preview for ChatGPT Business, Enterprise, Edu, and Teachers. Usage can draw from included allowances or workspace credits, with the actual amount depending on model, context, tools, caching, output length, surface, and plan. Public pricing guidance has changed more than once, so teams should review their current workspace rate card rather than budget from launch examples.
Workspace Agents do not replace source-system permissions, records, or human accountability. Analytics and compliance logs also have documented scope limits; do not assume every file, action, tool call, or approval is captured in one universal audit stream.
Practical Tips
Start with one workflow that has a named owner, a stable source of truth, a clear output, and one explicit approval boundary. Separate drafting from official writeback, keep connected credentials narrow, and test failure paths before scheduling or Slack deployment.
Use ChatGPT Work for substantial personal tasks that may later reveal a repeatable pattern. Promote the workflow into a Workspace Agent only when the team needs shared ownership, governance, analytics, or automatic runs.
Verdict
ChatGPT Workspace Agents are a credible shared-agent layer for repeatable team operations. They are most useful when the process is already understood and the organization is ready to manage permissions, credits, approvals, and ownership as deliberately as the prompt itself.