Claude Cowork vs Chat: What the Merge Changed

Anthropic shipped the merge on September 16, 2026, and the question people put on X within the hour was blunt: which one do I use now? For most of this year the answer to Claude Cowork vs chat was a tab you picked before you typed anything. That tab is on its way out. The difference it stood for isn't.

The official post says the reasoning plainly: people used both, and the annoying part was deciding where a task belonged. So Anthropic stopped asking. Claude now looks at what you've described and routes it, and everything Cowork could do is reachable from any conversation.

Key Takeaways

• The merge started rolling out September 16, 2026 on Pro and Max plans across web, desktop, and mobile. Team and Free follow "soon" with no date attached; Enterprise admins get at least 30 days' notice.
• There's nothing to switch on, and once your account moves over you can't switch back to separate Chat and Cowork options.
• Your chats, tasks, projects, connectors, skills, artifacts, and files carry over. Old Cowork tasks still open the way they did.
• Five things are broken or missing right now, including Add from GitHub and conversation branching.
• Claude Docs and Claude Slides launched the same day, both in beta on paid plans.

Claude Cowork vs chat: the difference that outlived the tab

Claude Academy's tutorial on choosing between them is still up, and Anthropic put a banner across the top of it rather than deleting the page: Claude Cowork is now just Claude, and what you learn there still works. That's the most useful sentence anyone has written about this change. The two modes were never two products in any deep sense. They were two ways of handing work over.

Chat is a conversation you steer turn by turn. You're present for every step, the task is light on context, and what comes back is text you'll read or paste somewhere. Cowork was a delegation: you describe the outcome once, Claude works across your files and tools, and you come back to something finished.

So the honest answer to Claude Cowork vs chat, after September 16, is that the split survived the merge intact. What changed is who decides. You used to declare it up front by clicking a tab; now Claude infers it from how you phrase the request, which means the way you brief it carries more weight than it used to. Anthropic's own guidance leans into this, asking you to name the outcome you want, the format you want it delivered in, and the files or apps it should work from.

And that last bit is where the seam shows. Claude can now produce a report and a deck from one conversation, and they'll live at a shareable link. Somebody still has to carry the result into the place the work actually continues: the Notion page the team reads, the Airtable base tracking it, the Slack channel where it gets argued about. That handoff is the gap Zentor was built to close. It runs the job and puts the finished thing back in the tool the work already lives in, instead of handing you an artifact and a link to move yourself.

How to tell whether your account has switched

Anthropic is rolling this out in stages, and two accounts on the same plan will get it at different times, so "is it live" isn't a yes-or-no question about the product. It's a question about your account.

The test takes two seconds. Look at your message box: if it still offers you Chat and Cowork as options, you don't have the new experience yet. Keep working the way you have been. When your account moves over, your chats, tasks, projects, and settings come with it.

One more wrinkle for anyone supporting a team through this: because the staging is per account rather than per plan, two people sitting next to each other can be looking at different products while describing the same one. Half the confusion in the first day of reports came from exactly that.

Worth knowing before you go looking for a toggle: there isn't one, in either direction. You can't opt in early, and you can't go back once it lands. Anthropic's help page says so in a single line, and that line is the whole migration policy.

What moved, and where to find it

Most of the anxiety around this change is misplaced. Nothing was deleted. A handful of things sit somewhere new, and if you built habits around them you'll want to check each one.

Cowork tasks now sit in Recents alongside your chats, in one list rather than two. Your Cowork Global instructions setting folded into Instructions for Claude under Settings, General, and Anthropic explicitly asks you to go read yours and confirm they still say what you want, which is a reasonable hint that the merge of two instruction sets is the part most likely to surprise someone. The web search toggle is gone entirely; Claude searches when it judges that searching helps. Deep research moved to a slash command, /deep-research, or the plus button under the chat input. Files Claude creates now appear next to the conversation, and on Claude Desktop the folder Cowork used to save into is still there under Settings, Storage folder, with any folders you'd granted access to listed under Trusted folders.

The model picker and the Code tab didn't move at all.

If you'd written internal documentation that walks colleagues through Claude Cowork vs chat as a choice they make before starting, that's the page to revise this week. The advice inside it is still sound; the first instruction in it is now wrong.

One item on that list is quieter than the others and deserves a second read: memory from Cowork tasks that ran only on your computer stays with those tasks. Search in the new experience doesn't cover older Cowork tasks either, though you can still find them by name in Recents. So the context your assistant built up doesn't travel as one thing; it's partitioned by where the work happened to run. Zentor takes the opposite position on that, deliberately: what it remembers across tasks belongs to you and exports cleanly, so a product decision made in some vendor's release notes doesn't quietly strand half of it.

What doesn't work yet

Anthropic published the gaps rather than burying them, which is more than most shipping notes manage. As of September 17, 2026, five things are missing from the new experience:

• Add from GitHub isn't supported.
• Branching a conversation from an earlier point isn't available.
• Incognito chats still work, but they open in the old experience, so Claude can't create files or run code inside them.
• Search doesn't include older Cowork tasks.
• Dispatch is closed to new users; if you already use it, you keep it for now.

There's also a billing wrinkle nobody will notice until they do. Everything counts against your plan's usage limits, and longer agentic work that searches, runs code, or builds files eats more than a quick question. While the rollout is in flight, Anthropic says usage "may be measured slightly differently" between accounts that have the new experience and accounts that don't. If your consumption looks strange this month against the same workload, that's a plausible reason before you go hunting for one. Settings, Usage still shows where you stand.

None of this is a botched release. It's the ordinary cost of a single vendor owning the whole surface: the shape of your workspace changes when they decide it changes, and your only move is to relearn it. That's the structural argument for keeping at least some of your recurring work somewhere the output is portable, which is the design Zentor starts from rather than the one it grew into.

Where the finished work actually lands

Read the capability list closely and there's a line that matters more than the headline feature. Longer tasks keep running in the cloud after you close your laptop or navigate away, and the result is waiting in the conversation when you come back. But tasks that touch files or apps on your computer need Claude Desktop open.

That's a sensible engineering boundary, not a flaw. It does mean the class of work most knowledge workers care about, the kind that reaches into a real folder or a real application, still has your machine in the chain. Your laptop's state becomes part of whether the job finishes.

This is the one place where the comparison to Zentor is concrete rather than rhetorical: the run doesn't depend on your machine being awake, and the output doesn't stop at a link. Both matter more for a recurring task than for a one-off. If the weekly report is genuinely going to write itself every Monday, the thing that decides whether it's useful is where it ends up on Monday morning, not how good the draft was.

For a wider view of how these assistants compare on delivery rather than on model quality, we walked through Grok Bot, Claude Cowork, and ChatGPT Work side by side. If you're on Windows and Cowork was already misbehaving before this change, the Windows-specific failures are a separate problem with separate fixes. And for the patterns that hold up regardless of which tab exists, there's our set of recipes for knowledge workers.

FAQ

Is Claude Cowork discontinued?

No, and the framing of Claude Cowork vs chat as two products was always looser than it looked. Anthropic merged it into the main Claude experience rather than retiring it. Everything Cowork did is available from any conversation, and your existing Cowork tasks, projects, connectors, skills, artifacts, and files carried over. The help documentation still describes the Cowork way of working, because the way of working didn't change; only the place you start it did.

Where did my old Cowork projects go?

They're in Recents, mixed in with your chats rather than kept in a separate list, and projects themselves now work in every conversation. On Claude Desktop, the folder Cowork saved output into is unchanged and lives under Settings, Storage folder. Folders you'd previously given Cowork access to appear under Trusted folders.

Do I need to turn anything on?

No, and you can't. The rollout is automatic and staged, so accounts on the same plan land at different times. If your message box still shows Chat and Cowork as separate options, it hasn't reached you yet. Once it has, switching back isn't an option.

When do Team and Free plans get it?

Anthropic hasn't given a date, only "soon." Pro and Max went first on September 16, 2026. Enterprise organizations are on a different track and their admins get a minimum of 30 days' notice before anything changes.

What are Claude Docs and Claude Slides?

Both launched alongside the merge on September 16, 2026. Ask for a document or a presentation and Claude drafts it inside the conversation, and you can edit it directly, present from Claude, or export to PowerPoint or PDF. Claude Design works inside conversations now too. All three are in beta on paid plans, with Enterprise admins choosing when to enable them. Standalone Claude Design keeps working the way it did.