OpenAI Launches Workspace Agents in ChatGPT, With a Free Window Closing May 6
OpenAI launched Workspace Agents in ChatGPT on April 22, 2026. These are shared, cloud-hosted agents built on Codex, aimed at team workflows instead of one-on-one chats. Right now, anyone on Business, Enterprise, Edu, or Teachers plans can try them for free. Pricing kicks in after May 6.
What Workspace Agents actually do
Workspace Agents are the next step after GPTs. A GPT is just a tuned chat partner. A Workspace Agent runs in the cloud, handles multi-step workflows, and can be shared across a team. OpenAI says agents can write and run code, connect to outside tools, remember things between sessions, run on a schedule, and answer requests in Slack.
The memory model looks simple: each agent gets a file store it can read and write as needed. OpenAI hasn’t published limits on file size, tool call depth, or how much context an agent can keep in scope. The details are missing from the announcement.
OpenAI gives five examples: software request reviewer, product feedback router, weekly metrics reporter, lead outreach agent, and third-party risk manager. Internally, their accounting team uses an agent for month-end close, and sales uses one to pull call notes and draft follow-ups.
Permission model and governance controls
Permissions are layered. Builders pick which tools and data an agent can reach, and can require human approval for sensitive actions like sending emails or editing spreadsheets. Enterprise and Edu admins get more control: they can set role-based access for tools and suspend agents if needed.
Admins get a Compliance API log for every agent’s config, updates, and runs. OpenAI claims built-in defenses against prompt injection, pointing to its safety research. An admin console view is coming that will show all agents, usage, and connected data, but it’s not live yet.
The announcement skips some key details: no hard limits on tool calls per run, no clarity on how memory works when several people use the same agent, and nothing on data residency or retention for agent output. If you’re in a regulated industry, check the enterprise data agreements and Help Center before trusting it with sensitive work.
Integration boundaries matter
Right now, you can only run Workspace Agents from ChatGPT or Slack. There’s no API trigger yet. OpenAI says automatic triggers and Codex app support are coming, but there’s no date.
That’s a real limit for anyone wanting to embed agents in their own tools. Until API triggers arrive, Workspace Agents only work inside ChatGPT and Slack. They’re not a general-purpose agent runtime yet.
What the community is asking
A Hacker News thread surfaced several practical concerns that the announcement leaves open. One commenter noted that the announcement gives “only a vague idea about what this is” and called for clearer documentation of the execution environment and available tools. Another raised the data privacy question directly: “Sending your entire communication and documents to OpenAI would be a very bold choice”.
There’s a bigger issue: data quality. If your internal docs are a mess, agents will surface that fast. Stale Confluence pages and inconsistent sources mean unreliable results, no matter how good the model is. Workspace Agents don’t cause this, but they will make it obvious.
Notion already has shared custom agents. Google just launched an agents CLI for its enterprise platform. OpenAI’s edge is distribution: ChatGPT Business and Enterprise are already in use, and Slack integration means non-technical teams can try agents without changing their workflow.
Where to find documentation before May 6
OpenAI directs teams to two resources: The Academy course on Workspace Agents and the Help Center article. The Help Center is where you’ll find the real configuration details that didn’t make it into the announcement.
What to watch before committing workflows
If you’re thinking about trying Workspace Agents before the free window ends, here’s what to do first:
- Check the Help Center docs on tool-call limits, memory scoping, and data handling before you build anything with sensitive data.
- Start with low-risk workflows. The research preview label means there are still rough edges.
- Ask your legal or compliance team if the Compliance API log is enough for your audits, especially if you’re in a regulated industry.
- Watch the API trigger roadmap. If you need to embed agents outside ChatGPT or Slack, you’ll have to wait as there’s no ship date yet.
- Audit your data sources. Agents are only as good as the content they can reach. Stale or inconsistent docs are a known failure mode.
The May 6 pricing date gives teams roughly two weeks to evaluate the feature. Teams have about two weeks to try Workspace Agents for free before pricing starts on May 6. That’s enough time for a proof of concept on a small workflow, but not enough to test governance for anything critical.
Sponsored
Claude Design
Claude Design turns conversation into polished prototypes, slide decks, and one-pagers. Describe what you need, Claude builds a first version, and you refine through inline comments, edits, or sliders — kept on-brand via…
View tool
Comments
No comments yet
Be the first to share your thoughts