Streamlines AI workflows for Pro and Max subscribers by combining standalone agent tasks with regular chat.
What is the Claude Cowork and chat merge?
Anthropic merged Claude Cowork and standard Claude chat into a single general agent on September 16, so quick questions and long-running delegations now live in one interface. The company’s announcement puts it in one line: bring a quick question, or hand over a report due at noon, and Claude takes it from there, even after you close your laptop.
The rollout reaches Pro and Max plans first across web, desktop, and mobile over the coming weeks, for existing and new users on those plans, with more plans to follow. Cowork users keep their existing chats, projects, artifacts, connectors, and skills where they left them.
The same announcement launched Claude Docs and Claude Slides, expanded Claude Design, and put all 3 into beta on paid plans, with Enterprise admins choosing when to switch them on. Anthropic’s framing is that Claude figures out what a task needs instead of making you choose the surface first.
Claude is now one surface for asking and delegating, and the asking side keeps its guardrails only if you configure them.
Does the merged Claude agent actually work?
The announcement confirms the feature exists and ships now, and it includes no performance benchmarks for the merged agent. Anthropic’s positioning leans on delegation rather than measured output quality, so the honest answer is that the proof is not published yet.
Simon Willison, who covered the merge the day it dropped, expects the boundary-mapping to take a while: figuring out what the merge means for features and surfaces is still going to take quite a bit of work, in his words. He reads the move as Claude becoming a general agent in its own right.
One control is documented: by default Claude asks before taking an action, you can grant it more autonomy, and Anthropic states you keep the final say. That permission model is the most buyer-relevant spec in the announcement, and it is worth testing before any long-horizon work goes through.
The capability is real and shipped, the benchmarks are absent, and the permission default is the part to verify yourself.
Claude Cowork vs chat: which one should you use now?
Before September 16 the answer depended on the job: chat steered a conversation turn by turn, and Cowork ran delegation across files and longer horizons. After the merge the question dissolves, because both behaviors run from any conversation with your existing context, skills, and connectors attached.
The comparison that still matters sits at the vendor level: Willison points at the echo with OpenAI’s ChatGPT Work push, which puts Claude and ChatGPT on the same general-agent path from opposite directions. For a buyer, the practical contrast is between one surface with a documented permission default and a stack of separate tools you stitch together yourself.
The Cowork-versus-chat decision is gone inside Claude, and the autonomy setting is the decision that replaced it.
The client deck is due at noon and the school pickup does not move, so the laptop lid closes with the work still running. Under the old split, that handoff started with a sorting question: quick chat or Cowork project, and picking wrong meant rebuilding the task in the other window.
The merge deletes that sorting step for Pro and Max subscribers, the $20 and from-$100 plans where solo founders and small teams already live. Delegating a noon deadline to an agent that keeps working after the laptop closes buys back hours you do not have.
The multiplier has a tail, because Docs, Slides, and Design betas shipped on paid plans the same day, which means Claude renders and formats a deck while you drive. Your autonomy setting is the difference between a convenience and an incident.
Who is the merged Claude agent for?
The clearest fit is a Pro or Max subscriber who already mixes quick queries with multi-hour deliverables every week, and a founder or senior lead handing off bounded tasks sits at the center of it. The noon-report pattern in the announcement is that workload.
Free-tier users see no change this quarter, because the rollout names Pro and Max first with more plans to follow and no dates attached. Teams on Enterprise plans wait on admin controls, since admins choose when Docs, Slides, and Design turn on.
We track how agent vendors handle handoff and autonomy in our AI signal tracker, and this merge is the sharpest surface consolidation in the agent market since OpenAI folded its coding agent into ChatGPT.
If your week already mixes quick questions and long deliverables in Claude, the merge removes a step you were paying for in attention.
Is Claude worth $20 a month for agent work?
Pro costs $20 a month billed monthly, or $17 a month on the annual plan at $200 up front, and the merge adds hand-off, scheduling, Docs, Slides, and Design to that price with no stated surcharge. Anthropic’s pricing page lists Max from $100 a month for heavier usage.
At $20 the test is cheap for anyone already paying for 1 AI subscription: run a recurring low-stakes task, like a weekly report draft, through the merged agent and watch how it behaves across a few runs before anything client-facing depends on it. If you sit on the free tier, the merge changes nothing for you this quarter.
Migrate your lowest-stakes recurring task this month and let it fail cheap before anything important depends on it.
Source: simonwillison.net
