Anthropic is gradually merging Claude Chat and Cowork for Pro and Max users, so an ordinary conversation can become a longer task that reads files, runs code or produces an editable artifact without first switching products. The change matters because the chat thread is becoming the durable unit of work, but it does not erase the permission, local-device and storage boundaries behind that interface.
Key facts
- The unified experience is rolling out gradually to Pro and Max users.
- Team and Enterprise keep separate Chat and Cowork surfaces for now.
- Manual permission mode is the default; Auto can continue with automated safety checks.
- The primary source is Anthropic's “Cowork is now Claude” announcement.
Anthropic's product claim is simple: a person can ask about a sales pipeline, ask Claude to write a report in the team's style, then ask for a five-slide presentation in the same conversation. The system can use conversation context and work across web, desktop and mobile. Anthropic calls the direction “one Claude,” a useful phrase for a UI merger but not evidence of one uniform execution environment.
The migration details are the real story. Anthropic's current help guidance says chats, Cowork tasks, projects, settings, connectors, skills, artifacts and files carry over. Yet older Cowork tasks do not appear in search; users find them by name in Recents. Global Cowork instructions move into general Instructions for Claude. Memory from tasks that ran only on a local computer stays attached to those tasks. Incognito chats stay in the older experience and cannot create files or run code. Local files, the built-in browser and computer use still require Claude Desktop to be open.
That makes the right analogy a hotel lobby with one reception desk. The guest no longer chooses a separate building at check-in, but the kitchen, laundry, key system and staff permissions still differ behind the desk. The visible mode selector disappears; the capability boundary remains. This is particularly important for administrators evaluating where a task runs and which data controls apply.
The merger arrives alongside increasingly concrete output surfaces. Claude Docs is a collaborative artifact with headings, tables, comments and export to docx, PDF, Markdown and Google Docs. Claude Design uses HTML in a chat/canvas workflow for mockups, interactive flows and visual assets; it can export ZIP, PDF, PPTX and, in the stated surface, Google Slides. These are usable handoff formats, not merely chat responses.
The strongest counterargument is that simplifying the composer may obscure whether Claude is about to take cloud actions or reach local resources. Anthropic's own rollout caveats validate that concern. Team and Enterprise administrators still have separate controls, and Enterprise is promised at least 30 days' notice before the merger changes an organization.
The so-what is not “agents have disappeared into chat.” It is that conversational context is being promoted from transient interface state to a work orchestration layer. Buyers should test the policy path of a real task—what it can access, where it executes, how it persists and how it is shared—rather than treating a unified title as a unified risk model.
Originally published on Ground Truth, where every claim is checked against the primary source.
Top comments (0)