Claude Enterprise unifies chat and Cowork with role access and usage controls

Anthropic has [made its unified chat and Cowork experience generally available for Claude Enterprise plans](https://claude.com/resources/articles/unified‑claude‑enterprise). The October 23 announcement extends an experience that reached Pro and Max users in September: a conversation can begin with a simple question, then continue into work that requires code execution or files without the user switching between separate chat and Cowork experiences. Anthropic says Claude can return a document, analysis or presentation from that cloud‑based work. The experience spans web, desktop and mobile, with existing chats, projects, connectors, skills and plugins carrying over. For a team, the useful change is continuity between a question and a larger assignment, rather than having to decide which interface to use before starting. This rollout is separate from [Claude Duet’s research preview](https://opentools.ai/news/claude‑duet‑preview‑waitlist‑background‑usage‑approval‑memory), which begins with a US Pro and Max waitlist. The Enterprise announcement concerns unifying chat and Cowork, rather than opening Duet to all workplace accounts.

[Agentic work runs in an environment Anthropic manages](https://claude.com/resources/articles/unified‑claude‑enterprise), rather than depending on a user keeping their laptop open. Users can leave a longer task running and pick it up from another device. That can make it easier to hand over a project before stepping away, but the announcement does not establish a completion‑time guarantee or a typical productivity gain. For Enterprise cloud sessions, Anthropic describes isolated virtual machines, network egress switched off by default and connector credentials kept outside the sandbox. It also says the unified experience builds on existing controls including single sign‑on and the Compliance API, with data encrypted in transit and at rest and not used for model training by default. These are the company’s architecture and policy statements; the release post is not an independent security assessment.

[Claude Enterprise billing remains usage‑based](https://claude.com/resources/articles/unified‑claude‑enterprise). Anthropic cautions that agentic tasks can use more tokens than a quick chat. Bringing both kinds of work into one conversation therefore does not make their costs equivalent. Admins can use model entitlements to determine which models each role may access. A new Efficient mode keeps the unified experience on Sonnet and Haiku models. Spend limits can be set at organization, group and member level, while pooled group budgets provide a shared monthly amount for a group to draw from. These controls help define a rollout’s boundaries; they do not establish what a particular organization will spend or save. A practical starting point is to identify which roles need extended project work, decide their model access and set the relevant budgets before expanding usage. A familiar chat interface can now lead to a larger cloud task, so the work assigned matters alongside the number of conversations.

The [rollout announcement](https://claude.com/resources/articles/unified‑claude‑enterprise) explicitly says organizations using zero data retention for Claude Code, a HIPAA‑ready configuration or customer‑managed encryption keys stay on their current configuration. General availability should therefore not be read as an automatic migration of every existing Enterprise organization. Admins can manage the rollout in Organization settings, and Anthropic says existing Enterprise customers will receive an email about their organization’s rollout. For a team planning adoption, the first checks are its actual configuration and the access settings available to its admins. The new experience changes how supported work starts and continues, while eligibility, permissions and usage remain decisions that need deliberate setup.