OpenAI is folding the separate Chat and Work tabs in the ChatGPT desktop app into one experience. According to TechCrunch's report on an email OpenAI sent to users, the company will first offer the merged version to select users for feedback and then roll it out widely. No dates were given. Codex stays a separate experience, and ChatGPT and Codex will draw from one shared token allowance.
If you have been confused about where to start a task, you are the audience. This post explains what exists today, what the merge changes, what stays apart, and what to watch if you pay for a plan.
TL;DR: what is confirmed
| Question | Answer |
|---|---|
| What changes? | Chat and Work tabs in the desktop app become one experience |
| What does the email promise? | Move from exploring an idea to having ChatGPT carry it through, in one place |
| What can it do? | Add files and plugins, then ask for a website, a presentation or web tasks, while you review the work |
| Does Codex merge? | No. It "will remain a separate experience in the ChatGPT desktop app" |
| When? | Select users first for feedback, then wide rollout. No dates |
| Limits | ChatGPT and Codex share one token allowance; most users unaffected, some may wait for resets more often |
| Source | OpenAI email to users, as reported by TechCrunch |
What are the Chat and Work tabs today?
The current layout comes from OpenAI's new desktop app. OpenAI's help article on moving to the new ChatGPT desktop app says the app now includes Chat and Work under ChatGPT, alongside Codex, on macOS and Windows. It describes the three parts this way:
- Chat: ask questions, search and have conversations.
- Work: research, analyze information, and create documents, spreadsheets, presentations, reports and Sites.
- Codex: develop software with local files, repositories, terminals and developer tools.
On desktop you pick ChatGPT or Codex from a menu. Inside ChatGPT you flip a toggle between Chat and Work. Chat and Work chats appear together in Recents, and projects let you start either kind from the same context. The older app survives as ChatGPT Classic, which OpenAI says keeps receiving model updates, bug fixes and security patches.
So the split today is a toggle inside one app. TechCrunch notes that the app was already confusing, with ChatGPT and Codex sections and two tabs within ChatGPT. The merge removes the second layer of that choice.
What does merging Chat and Work actually mean?
The email, as quoted by TechCrunch, states that the change "will enable you to move from exploring an idea to having ChatGPT carry it through, all in one place." The practical meaning is that you no longer decide in advance whether a request is a conversation or a task.
Under the old toggle, a question such as "what should my pricing page say?" belonged in Chat. A request such as "build the pricing page and a deck for it" belonged in Work, where the agent can create files and use plugins. The new flow lets you start with the first and continue into the second without switching tabs or re-explaining the context.
According to the report, you will be able to add files and plugins and ask ChatGPT to build a website, create a presentation or complete web tasks, then review its work. That is the Work capability, now reachable from the main chat surface.
We have not tested the merged build, because it is not yet generally available. Treat the description as OpenAI's plan. We will update this post when the rollout reaches us. For a feature-by-feature account of the agent side, see our ChatGPT Work vs Codex guide.
What stays separate: Codex
The email is explicit: "Codex will remain a separate experience in the ChatGPT desktop app." OpenAI's help article agrees. It says Codex remains a separate view, its workflows are unchanged, and its history remains separate from ChatGPT history. Codex is also not selectable on web or mobile, though supported desktop Codex chats can be reached from the Remote tab in the ChatGPT mobile app.
The reason is the work itself. Codex runs against local files, repositories and terminals, with permissions and sandboxing that a chat agent does not need. Keeping it separate preserves those boundaries. It also keeps the coding agent from inheriting the casual defaults of a chat window.
The shared allowance: the part that touches your wallet
The merge arrives with a limits change that matters more to paying users than the tab layout. OpenAI says ChatGPT and Codex will draw from the same token allowance. TechCrunch reports that most users will not be affected, but some may need to wait for usage resets more often, depending on their usage and the model they choose.
That matches the notice we covered in detail in our post on the shared ChatGPT and Codex token allowance starting in December. The confirmed facts there: Plus, Pro and Business plans, usage measured in tokens, Instant chat not counted, and a combined view under Settings, then Usage. What is still missing is any per-plan token number.
Here is what that means in practice:
| If you mostly... | What to expect |
|---|---|
| Chat in short sessions | Likely no change |
| Run long Work tasks that create documents or sites | Those tasks draw from the pool; heavy runs may exhaust it faster |
| Use Codex heavily plus ChatGPT | One pool now covers both, so a long coding session can leave less for chat |
| Pick the most advanced model for everything | Larger models use more allowance per task |
A related change affects who can use what, and where. We covered the per-device limits on paid accounts earlier this month. Together they show OpenAI tightening how plans are measured while it simplifies the interface.
How does this compare with what Anthropic did?
TechCrunch notes that Anthropic merged Claude's chat and Cowork functions into one screen in September. We wrote about that change in our Claude Cowork chat merge post. The direction is the same: the line between answering and doing is moving out of the navigation and into the model's judgment.
| OpenAI (announced Nov 2) | Anthropic (September) | |
|---|---|---|
| Chat and agent work | Merging into one experience | Merged into one screen |
| Coding agent | Codex stays separate | Handled as its own surface, per our earlier post |
| Rollout | Select users, then wide; no dates | Described in our September post |
| Limits | Shared ChatGPT and Codex allowance | See Anthropic's own plan pages |
We keep the comparison to what each company has said. Neither has published a head-to-head on how often their agents finish tasks without a human fix, so any claim that one merge is better is not yet supported.
What this means for builders and teams
- Stop planning around tabs. If you wrote internal guides that say "use Chat for X and Work for Y," plan to rewrite them when your account gets the merged build.
- Watch your usage page. With one pool, track combined remaining usage before launching a long agent run. A long task can use more than a long conversation.
- Review permissions as agent reach grows. When one window can chat and act, the question of what it may touch matters more. Our guide to ChatGPT controlling your Chrome tabs lists the risks. If you run agents with real access, AgentBeam, the agent security platform from the explainx.ai team, is built to stop AI agents before they take dangerous actions.
- Plan for outages as one surface. More features behind one door means one failure can affect more of them. See what happened in the August 31 ChatGPT Work outage.
- Keep Codex habits. Nothing about your coding workflow changes under the announcement.
What we know and what we do not
Confirmed: OpenAI emailed users about merging Chat and Work, with Codex staying separate and a shared token allowance, as reported by TechCrunch on November 2, 2026. The current three-part app structure is documented in OpenAI's own help center.
Not confirmed: the rollout date, which users get it first, whether the toggle disappears entirely or becomes automatic routing, whether Free users get the same experience, and how large the shared pool is per plan. We did not see the email itself, only TechCrunch's account of it.
Our read: the merge is a navigation fix with a pricing consequence attached. The interface will matter less than the usage page. If you pay for ChatGPT, check how much of your allowance a typical Work task burns before the December limits change reaches you.
Related reading
- ChatGPT and Codex shared token allowance from December
- ChatGPT Work vs Codex: the complete guide
- Claude Cowork and chat merge
- ChatGPT can read and control your Chrome tabs
- ChatGPT per-device card limits on paid plans
- ChatGPT Work outage of August 31
- Sources: TechCrunch report, OpenAI help center: moving to the new ChatGPT desktop app, OpenAI help center: ChatGPT and Codex usage limits
Details are accurate as of November 2, 2026. OpenAI may change the rollout and limits.