Cowork is now Claude: what teams need to know

On 16 September 2026, Anthropic announced the gradual merger of Claude Cowork and chat into a single Claude. If you knew Claude Cowork as a separate mode for bigger tasks, you no longer pick a mode after the switch: Claude decides for itself whether a request needs a quick answer or is a multi-step task (Anthropic blog, 16 Sep 2026).

For teams this is more than a new layout. The rollout is staggered by plan. For Pro and Max the place where tasks run also changes on 6 October, and on Enterprise plans new defaults for browser access have applied since 10 September. Here are the key points about Claude Cowork, summarized in English, each with an Anthropic source (as of 5 October 2026).

TL;DR

• Chat and Cowork become one Claude: no more switching modes, gradual rollout first on Pro and Max, Team and Free follow, Enterprise admins get at least 30 days' notice.
• Four settings matter for admins now: browser access (Claude in Chrome and the built-in browser), approval modes, running in the cloud and, on Enterprise, the new smart reports.
• Roll out in four steps: rules and approvals, small pilot, training, documentation of AI literacy measures.

What happened on 16 September

On 16 September 2026, Anthropic announced that Claude Cowork and chat are merging into one Claude. You describe what you need, and Claude decides which tool to use for it. According to the Help Center, with the new interface everything that used to be possible only in Cowork is available from any conversation. Anthropic itself names one exception: incognito chats continue to open in the previous interface, where Claude cannot create files or run code (Claude Cowork and chat are one Claude).

At the same time, Claude Docs and Claude Slides were added, and Claude Design now works directly in conversations. Results can be edited in place, presented from within Claude or downloaded as PowerPoint or PDF. According to the Anthropic blog, all three are in beta on paid plans (Anthropic blog, 16 Sep 2026). The release notes of the same day, by contrast, list artifacts including Design, Slides and Docs on every plan, Free included, and specify for Enterprise: there the three are in beta and off by default until an owner turns them on (Release notes, 16 Sep 2026).

This is how the rollout works according to Anthropic:

• Pro and Max: gradually over the coming weeks, on web, desktop and mobile. There is nothing to switch on.
• Team and Free: follow "soon" according to Anthropic, with no fixed date.
• Enterprise: admins hear from Anthropic at least 30 days before anything changes for their organization.
• Team and Enterprise today: both keep chat and Cowork as they are for now (Use Claude Cowork on Team and Enterprise plans).

By default, Claude asks before it performs an action ("Manual" mode). If you want, you can switch to "Auto": Claude then keeps working without asking at every step, and automatic safety checks run before every action. Chats, projects, artifacts, connectors and skills from Cowork move along; after the switch there is no way back to the separate chat and Cowork selection.

The step has a history. On 7 July 2026, Anthropic brought Cowork to web and mobile, starting with the Max plan: since then, sessions keep running in the cloud (beta) when the laptop is closed, and scheduled tasks do not need a device that is switched on (Release notes, 7 Jul 2026).

What changes for Pro and Max on 6 October

For Pro and Max plans, Anthropic announces a further change: from 6 October 2026, new Cowork tasks run in the cloud, and the "Only on your computer" option in the settings is removed (Use Claude Cowork on web, desktop, and mobile). Tasks that were already started locally stay on the computer and can be completed. Each one shows a notice with a download of the history in case you want to continue the work in Claude Code.

Your folders stay on your computer. Claude only reaches connected folders, through the desktop app and only while it is open. If a task needs a file in the cloud, Claude fetches a copy of exactly that file. If you delete the session, the retrieved copies are also deleted in line with Anthropic's retention rules. According to Anthropic, a deleted Cowork task disappears from the history immediately and is deleted from the backend systems within 30 days (Get started with Claude Cowork). Scheduled tasks also move to the cloud.

For anyone who has to keep work on a single computer, Anthropic itself names the alternative: Claude Code in the desktop app runs on your computer and keeps folders and history there.

Cowork, Claude Code and chat: who uses what

A practical question after the announcement: what about Claude Code, and when do you use which? In short, according to Anthropic, Cowork uses the same agent architecture as Claude Code, only without a terminal and for knowledge work beyond code (Get started with Claude Cowork). The table puts the three routes side by side, using only statements from Anthropic's own documentation.

Sources for the table: Claude pricing page, Team and Enterprise article on Claude Code, Get started with Claude Cowork, Claude Code overview and Claude Code security.

Claude Cowork vs Claude Code: the practical difference

From 6 October, the difference becomes especially tangible for Pro and Max. Cowork tasks run in an isolated environment on Anthropic's servers and reach your computer through the desktop app. Claude Code in the desktop app, by contrast, runs on your computer.

The line is no longer rigid, though. According to the release notes of 16 September, designs, presentations and documents can be requested in any conversation, including in Claude Code (Release notes, 16 Sep 2026). And a plugin you add is stored in your account and, according to the Help Center, works in chat and in Claude Code just as in Cowork (Get started with Claude Cowork).

For the Mittelstand this means: office teams start with Cowork, development teams with Claude Code. Two rollout paths do not mean two contracts, though, because according to Anthropic Claude Code is included in every seat of the Team plan.

What changes for admins and IT

The most important update for admins is, first of all, reassuring: on Team and Enterprise plans, chat and Cowork stay separate for now, and Enterprise admins get at least 30 days' notice. It is still worth looking at the organization settings now, because some defaults have changed.

Browser: Claude in Chrome and the built-in browser

In Cowork, Claude can work on the web in two ways: through the Claude in Chrome extension or through a browser built into the desktop app. The same allowlist and blocklist apply to both, and Anthropic recommends starting with a restrictive allowlist (Claude in Chrome admin controls).

• Claude in Chrome has been available in beta to all paid plans since 18 December 2025 (Release notes). On Enterprise the extension was off by default; from 10 September 2026 it is on by default, unless you deactivated it beforehand. If it is enabled on Enterprise, Anthropic does not automatically inform users.
• The built-in browser is on by default on Team plans. On Enterprise it was off at launch and, from 10 September 2026, is also on by default, unless you turned it off.
• The Chrome side panel runs as a Cowork session on Enterprise only once an admin turns on "Cowork in the cloud" and Claude in Chrome for the organization. Otherwise the classic side panel stays (Get started with Claude in Chrome).

Cowork settings you should know

These points are in the Help Center article for Team and Enterprise plans:

• Cowork on or off: Cowork is on by default, and owners can turn it off for the whole organization. On Enterprise it can be enabled for individual teams through groups and custom roles.
• Running in the cloud: the "Run Cowork in the cloud" setting is on by default on Team plans and off by default on Enterprise plans.
• Approval modes: the "Automatically approve" mode is allowed by default and can be turned off. "Always allow" for connector tools with write access is off by default; write actions are then approved per task.
• Logging: Cowork sessions via Claude, Claude Desktop and Claude Mobile end up in the Compliance API, and events can be streamed to SIEM systems via OpenTelemetry. Local sessions store the history on the users' computers, outside Anthropic's retention rules and without central deletion by admins.
• Zero Data Retention: not available for Claude in Chrome, "the same as Cowork", as Anthropic writes (Claude in Chrome admin controls).
• Microsoft 365: since 7 July 2026, the Microsoft 365 connector can also write (emails, calendar, files in OneDrive and SharePoint). Before the connector can perform write actions, a Microsoft Entra administrator has to approve the new permissions and an admin has to enable the write tools (Release notes).

Smart reports on Enterprise

Since 10 September 2026, smart reports have been available in beta on Enterprise plans. For this, Claude reads a sample of histories from chat, Claude Code and Cowork covering up to 28 days, groups them by work area, attributes costs, shows friction points and suggests recurring patterns as shared skills (Get started with smart reports). In the beta, up to 10 reports per month and organization are free. They are not available for CMEK or HIPAA configurations, and for Claude Code they are missing in Enterprise organizations that use Zero Data Retention.

Anthropic explicitly states that smart reports are not intended for evaluating individuals or for personnel decisions. The feature is off by default, as is the option to attribute sessions to individual users. Our recommendation: agree on these two settings with your data protection team and, if you have one, the works council before turning them on.

Rolling out Claude Cowork on Windows

Cowork runs in Claude Desktop for Windows on all paid plans and needs the latest version. For distribution, Anthropic provides MSIX packages for Intune, SCCM, group policy or PowerShell (Deploy Claude Desktop for Windows).

Two points often trip people up in practice: Cowork on Windows needs the Windows feature "Virtual Machine Platform", which only takes effect after a restart. And anyone who installs Claude themselves without administrator rights gets Claude, but no Cowork on the desktop.

Enable-WindowsOptionalFeature -Online -FeatureName VirtualMachinePlatform -All -NoRestart

Claude Code: pinning models and providers

For development teams, Anthropic expanded the managed settings in September (Claude Code changelog). With version 2.1.283 (25 September), new model versions stay blocked via availableModelsMatch: "exact" until an admin approves them, and deniedModels blocks individual models. Version 2.1.285 (29 September) uses allowedProviders to limit which API providers a machine may work with.

Which models are behind it now

In parallel with the merger, Anthropic launched a new model family. According to the Help Center, model selection stays where it was. Since 1 July 2026, Enterprise admins have been able to control which models and effort levels their users may use (beta) (Release notes).

Claude Opus 5.5 (22 September 2026): the first model of the 5.5 family. According to Anthropic, it works at the level of Claude Fable 5.1 on most tasks, costs 40 percent less to run than Opus 5 and produces output more than 30 percent faster. Via the API it costs 4 US dollars per million input tokens and 20 US dollars per million output tokens (as of October 2026). With the launch, Anthropic raised the five-hour limits on Pro, Max, Team and seat-based Enterprise plans (Introducing Claude Opus 5.5).

Claude Sonnet 5.5 (28 September 2026): the faster, cheaper complement. Anthropic positions Opus 5.5 for complex work that needs careful judgment, and Sonnet 5.5 for clearly scoped everyday tasks, bug fixing and finished documents, presentations and spreadsheets. Sonnet 5.5 costs 2 and 10 US dollars per million tokens, is more than 30 percent faster than Sonnet 5 according to Anthropic and up to 30 percent cheaper per task (as of October 2026, Introducing Claude Sonnet 5.5). Claude Haiku 5.5 is due to follow in the coming weeks.

Both models are available with Zero Data Retention. The API prices also matter for Enterprise plans that bill by usage: there, according to the pricing page, you pay a seat price plus usage at API prices.

A simple rule of thumb for everyday work in the team follows from this: reports, presentations and routine tasks fit the profile of Sonnet 5.5; difficult analyses and long, branching tasks fit Opus 5.5. Also note that, according to Anthropic, Auto mode in Cowork consumes more of the usage limit than the other modes.

Rolling out Cowork in your team in four steps

The merger makes Cowork visible to more people without anyone consciously choosing a mode. That is exactly why an orderly rollout is worthwhile before the new Claude arrives on your plan. How Cowork works in principle is shown in our Cowork explainer (as of June 2026, before the merger).

1. Set rules and approvals

Decide which data and folders Claude may see and which actions need approval. Anthropic itself advises caution when granting access to sensitive local files such as financial records, credentials or personal documents. Instead of broad access, Anthropic suggests a dedicated working folder for Claude, plus backups of important files. Anthropic strongly advises against managing sensitive information or performing actions with it through the built-in browser or Claude in Chrome (Use Claude Cowork safely).

Also decide whether "Automatically approve" stays allowed in your organization and which websites go on the allowlist. How much autonomy an agent should get per task is covered in our human-in-the-loop guide.

2. Pilot with a small group

Start with a few people and a few clearly scoped tasks. For Claude in Chrome, Anthropic describes a five-step pilot: enable the extension at organization level, set a restrictive allowlist, limit the installation through IT to the pilot group, share the safety guide and collect feedback (Claude in Chrome admin controls).

On Enterprise plans, you enable Cowork only for the pilot group through groups and custom roles. Smart reports then help you see where the work gets stuck, but for the team and not for evaluating individuals.

3. Claude Cowork training for the pilot team

Cowork requires different habits than a chat: describe the goal, check interim results, grant approvals deliberately. According to Anthropic, results get better when you specify three things: the desired outcome, the format and the inputs Claude should work with (Claude Cowork and chat are one Claude).

That includes recognizing anomalies. Anthropic advises monitoring the task as a whole rather than every command: if Claude accesses files or websites nobody mentioned, or the task grows beyond the assignment, stop the task.

Our license costs €599 per license, one-off, and includes the Cowork course and the Claude Code course. Details are on the pricing page.

4. Document AI literacy measures

Art. 4 of the AI Act requires measures to build AI literacy, but no specific level, no mandatory training and no certificate. The Federal Network Agency (Bundesnetzagentur) recommends documenting the measures well, and according to the European Commission, organizations can keep internal records of training and other measures for this purpose. What has applied since the Digital Omnibus is summarized in AI literacy under Article 4 after the Omnibus.

In practice that means: record who learned what in the pilot, which rules apply and when you will review them. Our AI policy template contains an evidence sheet for this. If you want to plan the rollout together, feel free to talk to us for 30 minutes.

FAQ

Is Claude Cowork the same as Claude Code?

No. According to Anthropic, Cowork uses the same agent architecture as Claude Code, but is built for knowledge work beyond code and needs no terminal. Claude Code is an agent for software development that runs in the terminal, in the IDE, in the desktop app and in the browser. For Pro and Max, Anthropic explicitly names Claude Code in the desktop app as the route for anyone who wants to keep tasks local.

Is Claude Cowork free?

No, according to the Help Center, Cowork is only included in the paid plans Pro, Max, Team and Enterprise. Pro costs 20 US dollars per month with monthly billing, a standard seat on the Team plan 25 US dollars per month or 20 US dollars with annual billing, each plus tax (as of October 2026, US prices on claude.com/pricing). Anthropic has announced that the new interface will also come to the Free plan later, but has not said to what extent.

Is Claude Code included in Claude Pro?

Yes. According to Anthropic, the Pro plan includes access to Claude Code, in the terminal and in supported IDEs such as VS Code and JetBrains. Claude Code is also included in every seat of the Team plan. API usage via the Claude Console, by contrast, is not included in the Pro plan and is billed separately.

Do Cowork and Claude Code share usage limits?

On Pro and Max, yes: according to Anthropic, the limits apply jointly to Claude and Claude Code, and everything you do with Claude counts toward your plan's limit. Multi-step tasks with web search, code or files consume more than a short question. On Team and seat-based Enterprise plans, admins can enable usage credit so that Claude, Cowork and Claude Code keep running after the limit is reached.

What happened to Claude Cowork? Is it going away?

Cowork is not going away: with the new interface, everything that used to be possible only in Cowork is available from any conversation, incognito chats excepted. According to Anthropic, existing Cowork tasks, projects, connectors, skills, artifacts and files are carried over, and the model selection and the Code tab stay where they were. On Team and Enterprise plans, chat and Cowork stay separate for now.

Is Claude Cowork safe for sensitive files?

Anthropic advises caution when granting access to sensitive local files such as financial records and describes the likelihood of an attack, despite all safeguards, as "non-zero". Claude may permanently delete files only with explicit permission. In cloud sessions, opened files are processed on Anthropic's servers, not only on your computer. How prompt injection works and what helps against it is explained in our prompt injection guide.

Sources

• Anthropic, Claude Cowork and chat are now one Claude (16 Sep 2026): https://claude.com/blog/cowork-is-now-claude
• Anthropic Help Center, Claude Cowork and chat are one Claude: https://support.claude.com/en/articles/16761823-claude-cowork-and-chat-are-one-claude
• Anthropic Help Center, Release Notes (entries 18 Dec 2025, 1 Jul 2026, 7 Jul 2026, 10 Sep 2026, 16 Sep 2026, 22 Sep 2026, 28 Sep 2026): https://support.claude.com/en/articles/12138966-release-notes
• Anthropic Help Center, Get started with Claude Cowork: https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork
• Anthropic Help Center, Use Claude Cowork on web, desktop, and mobile: https://support.claude.com/en/articles/15520349-use-claude-cowork-on-web-desktop-and-mobile
• Anthropic Help Center, Use Claude Cowork on Team and Enterprise plans: https://support.claude.com/en/articles/13455879-use-claude-cowork-on-team-and-enterprise-plans
• Anthropic Help Center, Use Claude Cowork safely: https://support.claude.com/en/articles/13364135-use-claude-cowork-safely
• Anthropic Help Center, Get started with Claude in Chrome: https://support.claude.com/en/articles/12012173-get-started-with-claude-in-chrome
• Anthropic Help Center, Claude in Chrome admin controls: https://support.claude.com/en/articles/13065128-claude-in-chrome-admin-controls
• Anthropic Help Center, Get started with smart reports: https://support.claude.com/en/articles/16893491-get-started-with-smart-reports
• Anthropic Help Center, Deploy Claude Desktop for Windows: https://support.claude.com/en/articles/12622703-deploy-claude-desktop-for-windows
• Anthropic Help Center, Use Claude Code with your Pro or Max plan: https://support.claude.com/en/articles/11145838-use-claude-code-with-your-pro-or-max-plan
• Anthropic Help Center, Use Claude Code with your Team or Enterprise plan: https://support.claude.com/en/articles/11845131-use-claude-code-with-your-team-or-enterprise-plan
• Anthropic Help Center, What is the Pro plan?: https://support.claude.com/en/articles/8325606-what-is-the-pro-plan
• Anthropic, Claude Pricing (as of October 2026): https://claude.com/pricing
• Claude Code Docs, Overview: https://code.claude.com/docs/en/overview
• Claude Code Docs, Security: https://code.claude.com/docs/en/security
• Claude Code Docs, Changelog (2.1.283, 2.1.285): https://code.claude.com/docs/en/changelog
• Anthropic, Introducing Claude Opus 5.5 (22 Sep 2026): https://www.anthropic.com/claude-opus-5-5
• Anthropic, Introducing Claude Sonnet 5.5 (28 Sep 2026): https://www.anthropic.com/claude-sonnet-5-5
• European Commission, AI Literacy Questions & Answers (as of 27 Jul 2026): https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers
• Bundesnetzagentur, KI-Kompetenz (AI literacy): https://www.bundesnetzagentur.de/DE/Fachthemen/Digitales/KI/7_Kompetenz/start.html