Claude Enterprise merges chat and Cowork ahead of a November default change

One conversation now covers chat and delegated work

Anthropic made its combined chat and Cowork experience generally available for Claude Enterprise on October 23. Employees can move from a question to work involving code and files within one conversation, with cloud tasks continuing across web, desktop and mobile. Administrators control who receives access.

The underlying merger began with Pro and Max on September 16. Anthropic had previously separated Cowork's larger assignments from chat, requiring users to choose where to begin. The Enterprise release extends that approach to managed organizations; it is not the first release of Claude's document or presentation capabilities.

Source 1Source 3

November 16 is a conditional default change

Existing organizations start with the setting off. On November 16, untouched settings turn on only when Chat and Cloud code execution and file creation are enabled. Keep disabled preserves an opt-out; earlier choices persist. Missing prerequisites delay activation until both are enabled.

Organizations using zero data retention, a HIPAA-ready configuration or customer-managed encryption keys remain excluded. They retain their existing setup, despite the broader general-availability announcement.

Access also remains subject to Claude's supported locations. Its help center lists markets including the United States, United Kingdom, Canada and Australia; the Enterprise announcement does not establish a separate worldwide rollout.

Source 1Source 2Source 10

The practical gain is continuity, with a local-computer limit

The earlier launch described turning a weekly business update into both a report and presentation within the same conversation. The useful change, in our analysis, is that the second output can build on the first task's context instead of requiring a separate handoff between modes.

For example, a hypothetical operations manager could ask Claude to compare supplied regional updates, then turn the findings into a spreadsheet and briefing. That illustrates the documented workflow, not a task we ran. Existing Cowork tasks, projects, connectors and skills carry over.

Cloud work can continue after a laptop closes, but tasks involving local files or applications need Claude Desktop open. Enterprise conversations default to Manual permissions, which asks before actions. The help center also lists missing conversation branching and Add from GitHub support; older Cowork tasks remain findable by name but are excluded from search.

Source 3Source 4

Role settings determine who can act

Anthropic describes isolated cloud virtual machines, sandbox network access disabled by default, and connector credentials held outside the sandbox. We have not independently assessed that architecture.

The role documentation explains a less visible deployment issue: custom permissions apply only after members move to Custom roles. Leaving someone on the built-in User role preserves organization-wide access. Connector grants also combine across roles, so blocking a connector in one role does not cancel permission granted by another.

That makes group membership consequential. In our reading, a restricted pilot depends on the permissions participants actually inherit, rather than simply creating a group with a restrictive name.

Source 1Source 7

A single interface still produces variable bills

The current price page lists Enterprise at US$20 per seat per month, billed annually before tax, plus consumption at API rates. The plan documentation distinguishes a 20-seat self-service minimum from 50 seats for sales-assisted purchasing. The seat charge buys access, with no included token allowance under the current usage-based model.

Existing contracts need separate treatment. Anthropic says older Chat and Chat + Claude Code seat pricing remains until renewal, when those organizations transition to the single Enterprise seat. Older Standard/Premium seat-based organizations retain their existing terms until migration and cannot continue that billing model beyond their next renewal. The interface announcement should not be read as an immediate rewrite of every contract.

Efficient mode limits the unified experience to Sonnet and Haiku; new sessions start with Sonnet 5.5 at low effort. Other products are unaffected.

Anthropic's consumption guide explains why controls matter: multi-step Cowork work generates intermediate token usage that users may not see. A group limit assigns a ceiling to each member; a pooled group budget, still beta, shares one monthly amount. Exhausting the pool pauses everyone in that group, and usage can slightly exceed a limit before stopping.

Administrators also get a separate analytics view for unified usage. The guide's example carries Beta labels and sample figures; those illustrate the reporting interface, not customer adoption or proof that the experience is enabled.

Source 2Source 5Source 6Source 9

Conversation history and retained outputs follow different rules

Anthropic's retention documentation sets a 30-day minimum for custom retention. Activity in a conversation or its associated cloud session resets the chat clock. Chats inside projects follow project retention instead, and projects are retained indefinitely by default. Merely choosing a short chat period therefore does not establish the same lifespan for project work.

Artifacts generally follow their own applicable retention period, rather than disappearing with their originating conversation; publicly linked artifacts have further exceptions. Scheduled tasks and their runs survive deletion of their originating conversation, so deleting a discussion does not stop recurring work.

Source 2Source 8