Claude Enterprise merges chat and Cowork, on by default from 16 November

On 23 October 2026, Anthropic made its unified chat and Cowork experience generally available for Claude Enterprise. Pro and Max plans got it in September. Chat and Cowork are now one Claude on web, desktop and mobile. A team can ask a quick question or hand over a full project in the same conversation.

Staff don't pick chat or Cowork before they start. Every conversation begins as a chat, and a simple question still gets a quick answer. When a task needs to run code or work with files, Claude carries on in a cloud environment Anthropic manages. It returns the finished document, analysis or deck.

The work keeps going after someone closes their laptop, and they can pick it up on web, desktop or mobile. Chats, projects, connectors, skills and plugins carry over. Admins manage the rollout in Organization settings, and existing Enterprise customers will get an email about it.

In admin settings the feature is called "Chat and Cowork unified". It is off by default in existing organisations and turns on by default on 16 November 2026. Until then, where no Owner has changed it, Organization settings > Capabilities says it will be enabled automatically on that date.

The Owner can choose Enable now or Keep disabled, and can still turn it off later. Where the setting was already turned on or off, that choice stays.

It needs the Chat capability and the Cloud code execution and file creation capability turned on. If either is off, no notice appears and nothing turns on, and after 16 November it turns on as soon as both are. It doesn't need Cowork or Cowork in the cloud, and Anthropic means it to replace both over time.

Turned on for the whole organisation, it reaches members with the User, Admin, Owner or Primary Owner role. Members on a custom role get it only if that role includes Chat and Cowork unified. New roles start with All capabilities, which includes it.

Anthropic advises moving anyone who shouldn't have it to a custom role with it turned off before switching on. Role membership can be synced from your identity provider through SCIM to add members in waves, and access is checked on every request.

Organisations using zero data retention for Claude Code, a HIPAA-ready configuration or customer-managed encryption keys (CMEK) stay on chat and Cowork as they are today. Their admins can't turn the new experience on. Anthropic says HIPAA support is expected after general availability, and CMEK support is planned for a later release. An organisation that turns on CMEK after turning on the new experience loses it.

By our count, 16 November is three weeks after the date we checked. An organisation that does nothing gets the new experience on that date, if both required capabilities are on.

The default reaches every member on the four standard roles. A company that wants agentic work for some teams only needs its custom roles in place before the date. New roles include the setting unless someone turns it off.

An Owner who picks Keep disabled keeps today's setup past 16 November, because an explicit choice stays. The organisation can then run Anthropic's recommended pilot of two to four weeks before everyone moves.

A regulated organisation on zero data retention, the HIPAA-ready configuration or CMEK has no switch to make yet and stays on today's chat and Cowork. Anthropic gives no date for HIPAA or CMEK support.

Anthropic says the experience builds on existing enterprise controls, including single sign-on (SSO) and the Compliance API. Data is encrypted in transit and at rest, and Anthropic doesn't use it for model training by default. SSO, SCIM and tenant restrictions apply unchanged, with no separate sign-in. Anthropic says running agentic work in one environment it manages gives the security team central visibility and controls.

Each conversation that needs one gets its own virtual machine managed by Anthropic. After a few minutes idle, the machine stops and its disk is saved as an encrypted snapshot. The sandbox expires 30 days after it was created, and connector credentials are kept outside it.

Network egress, the sandbox's access to the internet, is off by default for Enterprise. Admins can allow package managers only, a custom list of domains or all domains. Connector policies (Blocked, Needs approval, or set per role) are enforced by Anthropic outside the sandbox on every connector call.

When the "Skip all approvals" and "Automatically approve" modes are off, users must approve tool use, file edits and browsing. That applies to scheduled tasks and the built-in browser too. Sharing a conversation that used its sandbox needs both Share chats and Share cloud sessions on.

Through the desktop app, members can link their computers so Claude can work with local folders, local MCP servers and applications. Remote control and device policies govern that access.

Conversations are kept until the member deletes them, or as the custom retention setting says, with a minimum of 30 days. Cloud sessions started from a conversation, and their transcripts, are deleted with it. Files Claude gave in a conversation on or after 8 October 2026 are almost all deleted with it, unless they are used elsewhere.

Some data isn't deleted with a conversation or covered by custom retention. That includes Cowork cloud sessions a member starts directly, scheduled tasks and their runs, and Cowork history on a member's computer. The memory of members who have left, and transcripts and memory kept by Claude Code and other products, are outside it too. Scheduled tasks keep running after their conversation is deleted, until an admin turns off Scheduled tasks.

When a member moves to the new experience, their memory moves to a new memory system. The old memory may be deleted 14 days later, or sooner if it is past retention.

Claude Enterprise pricing stays usage-based. Anthropic says agentic work can use more tokens than a quick chat, depending on the task. Its pricing page lists Enterprise at a seat price plus usage at API rates. Seats are US$20 a month, billed annually, with a 20-seat minimum.

Anthropic also says that, billed through usage, Claude Opus 5.5 costs 40% less to run than Claude Opus 5 on typical workloads. It puts Claude Sonnet 5.5 at up to 30% less than Claude Sonnet 5 for most work.

Efficient mode, the first of four cost controls, keeps the unified experience on Sonnet and Haiku models. New sessions then start on Claude Sonnet 5.5 at low effort. Model entitlements set which models each role can use, in chat and agentic work alike. Spend limits work at organisation, group and member level, and pooled group budgets give a group one shared monthly amount.

Usage for members with access moves to a new reporting line, "Chat and Cowork unified", out of the Chat and Cowork lines. Anthropic says the change is to filing only, and counting and billing stay the same. New activity can take up to two days to show.

For a security team, network egress starts off for Enterprise, and connector rules are enforced outside the sandbox. The egress list, the two approval modes, sharing and Remote control are the settings that widen access, so review those first.

None of the three sources names a region for the sandbox. A company with data residency rules will likely need an answer from its Anthropic account team before it switches on.

A records or privacy team that relies on custom retention has gaps to cover. Scheduled tasks keep running after their conversation is deleted, and Cowork sessions a member starts directly sit outside the setting.

For a finance owner, the usage part of the bill can rise as more conversations turn into agentic work. Efficient mode and spend limits are the controls that hold it. A month-on-month report will also show Chat and Cowork usage falling unless it adds the new unified line.

By our arithmetic, the 20-seat minimum at US$20 a seat comes to US$400 a month, or US$4,800 a year billed annually, before any usage.