What is new?
Chat and Cowork were merged for the Pro and Max plans in September 2026. On 23 October 2026, Anthropic made the same unified experience generally available for Claude Enterprise. Employees no longer have to decide up front whether they are asking a quick question or handing over a whole project. Both happen in the same conversation, on web, desktop and mobile.
Every conversation starts as a chat. When a task needs code or files, Claude continues the work in the cloud, according to Anthropic, and returns the finished document, analysis or deck. Because the work runs in the cloud, it keeps going when someone closes their laptop. Chats, projects, connectors, skills and plugins carry over.
How does the cloud sandbox work?
According to the Help Center, every conversation that needs agentic work gets its own virtual machine managed by Anthropic. After a few minutes of inactivity it is stopped and its disk is saved as an encrypted snapshot. The sandbox environment expires 30 days after it was created.
Network egress from the sandbox is off by default for Claude Enterprise; connector credentials are kept outside the sandbox. Through the desktop app, employees can optionally link their computer so Claude can work with local folders, local MCP servers and applications. This is governed by the remote control and device policies.
What happens on 16 November 2026?
In existing organizations, the setting is called Chat and Cowork unified and starts switched off. According to the Help Center, it turns on automatically on 16 November 2026. Owners and Primary Owners see a notice under Organization settings > Capabilities and can choose Enable now to start earlier or Keep disabled to prevent activation. A decision already made stays in place.
Two capabilities are required: chat and Cloud code execution and file creation. If either is off, nothing turns on. Important for planning: activation applies to the User, Admin, Owner and Primary Owner roles. If you want to exclude certain people, move them to a custom role without this capability beforehand.
Which admin settings matter?
Existing enterprise controls continue to apply, such as single sign-on, SCIM and the Compliance API. According to Anthropic, data is encrypted in transit and at rest and not used for training by default. The Help Center lists the settings that matter most for the unified experience.
- Allow network egress: off, package managers only, a custom domain list or all domains
- Connector policies (Blocked, Needs approval, per role), enforced by Anthropic outside the sandbox
- Approval modes such as Skip all approvals and Automatically approve, remote control, trusted devices and scheduled tasks
- Sharing chats and cloud sessions: the more restrictive setting applies
How can costs be managed?
Claude Enterprise pricing stays usage-based. Anthropic notes that agentic work can use more tokens than a quick chat, depending on the task. Admins have four levers: the new Efficient mode, which keeps the unified experience on Sonnet and Haiku models, model entitlements per role, spend limits at the organization, group and member level, and pooled monthly budgets for groups.
In reports, the unified experience gets its own line. Chat and Cowork usage is then filed under Chat and Cowork unified. If you add up spend or tokens for chat and Cowork, the Help Center advises including this line, otherwise the numbers appear to drop.
What about retention and deletion?
Conversations are kept until employees delete them or the organization's custom retention period applies, with a minimum of 30 days. When a conversation is deleted, the cloud sessions started from it are deleted too, including transcripts, according to the Help Center. Not covered include Cowork cloud sessions started directly, scheduled tasks and Cowork history stored locally.
During the move, memory is transferred to a new memory system; according to the Help Center, the old memory may be deleted 14 days later. Data protection officers should add these points to existing documentation.
What should businesses do now?
Anthropic recommends sharing the article with IT, security and procurement, completing the security review, starting with a sandbox organization or a pilot group through a role, reviewing network, connector, sharing and remote settings and monitoring usage for two to four weeks before expanding the rollout in waves.
Decide deliberately before 16 November 2026 whether the setting should turn on automatically. Does this also apply to Team plans? The announcement names Enterprise; Pro and Max got the experience in September. The free AI System Check helps organize approvals and responsibilities across your team. These notes are suggestions, not customer results.
Keep it verifiable
