
Claude Cowork use cases follow the plugins Anthropic publishes for each team. Marketing drafts content, plans campaigns and reports on performance; sales researches prospects and prepares for calls; customer support triages tickets and drafts responses; engineering writes standup summaries and postmortems; operations documents processes and evaluates vendors; finance reconciles accounts and analyzes variances; and legal reviews contracts and triages NDAs. Whatever the team, it is set up in four parts: who gets Cowork, the team's plugin, connectors to the systems it works in, and a project that holds its files and instructions.
Claude Cowork is Anthropic's agent for knowledge work. It "uses the same agentic architecture that powers Claude Code, with no terminal required", runs each task on Anthropic's servers in an isolated environment, and reaches folders on your computer through the Claude desktop app, according to Anthropic's Get started with Claude Cowork. It comes with the paid plans: Pro, Max, Ultra, Team and Enterprise. A plugin "bundles skills, connectors, and sub-agents into a single package", and Anthropic publishes its team plugins as open source in its knowledge-work-plugins repository, so your team starts from a working plugin and adapts it to its own tools.
A Cowork session belongs to the person who runs it. Anthropic's Help Center says sessions "can't be shared with others", although the individual artifacts a session creates can be. What a team shares is the setup around those sessions. It comes in four parts, set in this order: access comes first, the plugin names the connectors it needs, and a project holds the files the connected work draws on.
On Team and Enterprise plans, "Claude Cowork is on by default, but organization owners can manually disable it", and in organizations that have enabled HIPAA it starts off, according to Anthropic's Use Claude Cowork on Team and Enterprise plans. On Enterprise plans owners "can use groups and custom roles to enable Cowork for specific teams", so marketing can start before finance does. Groups and custom roles are an Enterprise feature, so on Team plans the choice is the organization-wide switch.
The plugin is what turns Cowork into a marketing or a finance agent. Its skills hold what Anthropic calls "the domain expertise, best practices, and step-by-step workflows" Claude draws on, and its connectors link Claude to the team's tools through MCP servers. Owners distribute plugins through private plugin marketplaces and set each one as installed by default, available to install, required or hidden. A required plugin is installed for every member, and members "can't turn it off or remove it". On Enterprise plans admins can override these settings for specific groups. Build one plugin per team from Anthropic's starting point, with your templates, style rules and review steps written into its skills, and every member's session works from the same instructions.
Each of Anthropic's plugins names the connectors its team needs, from HubSpot and Klaviyo for marketing to Snowflake and BigQuery for finance. Write actions stay gated. Members approve each write-capable connector tool task by task, unless an owner turns on the setting "Allow 'Always allow' for connector tools", which is off by default. Anthropic's Team and Enterprise guide adds that "most custom connectors don't annotate their tools, so every tool on those connectors is gated". Where no connector exists for a system your team runs on, such as an in-house ERP, someone has to build one.
Projects "group related tasks into workspaces with their own files, context, instructions, and memory", and on the desktop app folder instructions add context for a local folder you select. Claude "can only read and write files in folders you've connected". Put the weekly report template, the brand guide or the close checklist in the project, and recurring work can then run as a scheduled task, which Anthropic says runs in the cloud "with no device online".
The seven teams below are the ones Anthropic's plugin repository and its announcement of 24 February 2026 describe in detail. Anthropic also publishes plugins for HR, design, product management and data.
Anthropic's marketing plugin is built to "draft content, plan campaigns, enforce brand voice, brief on competitors, and report on performance across channels", with connectors for Slack, Canva, Figma, HubSpot, Amplitude, Notion, Ahrefs, SimilarWeb and Klaviyo. Anthropic's own marketing example on its Claude Cowork page is a recurring task: "Every Monday morning, build last week's marketing readout." Write your brand voice and approved terms into the plugin's skills, connect the analytics tools the readout draws on, and name the person who approves anything before it is published.
The sales plugin researches prospects, prepares for calls, reviews the pipeline, drafts outreach and builds competitive battlecards, through connectors such as HubSpot, Close, Clay and ZoomInfo. The customer support plugin triages tickets, drafts responses, packages escalations, researches customer context and turns resolved issues into knowledge base articles, through Intercom, HubSpot, Guru and Jira among others. Both teams work from a record, the account or the ticket, so connect the CRM or the helpdesk before anything else, and decide which replies a person sends.
Anthropic builds Cowork "for non-coding knowledge work" and positions Claude Code for software engineering. The engineering plugin covers the work around the code: writing standup summaries, coordinating incident response, deploy checklists and postmortems. Keep your incident template and runbooks in the team's project and connect the tracker the team already uses. Plugins also work in Claude Code, so the team carries the same skills into the tool where its code is written.
The operations plugin handles "process documentation, vendor evaluations, change request tracking" and runbook creation. A status report built every week from the same inputs suits a scheduled task. Keep the report structure and the list of open projects in a project, let the task draft the report, and have the operations lead edit and send it. Budgets, headcount and vendor contracts pass through this team, so set the approval rules before the first run.
The finance plugin preps journal entries, reconciles accounts, generates financial statements, analyzes variances, manages the close and supports audits, with connectors for Snowflake, Databricks and BigQuery. The legal plugin reviews contracts, triages NDAs, navigates compliance, assesses risk and drafts templated responses, through Box, Egnyte, Jira and Microsoft 365, according to Anthropic's plugin repository. Both teams produce records an auditor or a counterparty may read later, so start them in Manual mode, described below, until the team trusts the output.
Owners on Team and Enterprise plans create plugin marketplaces and can stream Cowork events, including tool calls, file access and approval decisions, to your SIEM and observability tools through OpenTelemetry, which Anthropic says "doesn't replace audit logging for compliance purposes". The setting "Run Cowork in the cloud" is on by default on Team plans and off by default on Enterprise plans. For the compliance record, Claude Enterprise admins retrieve Cowork sessions through the Compliance API, which returns sessions from the most recent 30 days, according to Anthropic's guide for Team and Enterprise plans.
By default Anthropic does not use inputs or outputs from its commercial products, which it lists as "Claude for Work, Anthropic API, Claude Gov", to train its models, according to its Privacy Center; feedback you choose to send is the exception. Local session data on members' computers "is not subject to Anthropic's standard data retention policies", and admins cannot manage or delete it centrally. Anthropic also warns that "Cowork has unique risks due to its agentic nature and internet access", because Claude acts on the files and apps you connect.
Cowork has three permission modes. In Manual mode "Claude pauses and asks for approval for actions", so you approve each action. In Auto mode "Claude keeps working without stopping to ask about every step", and an automated check reviews each action first. In Skip mode "Claude doesn't pause to ask and nothing checks its actions automatically". Set the mode, the connected folders and the connector approvals team by team, starting with the teams that handle personal or financial data.
| Manual | Auto | Skip | |
|---|---|---|---|
| How much Claude asks | Claude pauses and asks for approval for actions | Claude keeps working without stopping to ask about every step | Claude doesn't pause to ask |
| Checks | You approve each action | An automated check reviews each action first | Nothing checks its actions automatically |
Start with one team and one workflow that runs by hand every day, such as the weekly status report or ticket triage. Ovidius connects Cowork to the systems your teams already use, such as Jira and your CRM, builds the connectors your systems need when none exist yet, and sets which actions wait for a person's approval. A rollout runs through architecture and security scoping, a connector build tested in staging on test data, approval rules calibrated with the people who will use them, and a launch with every team trained on its workflow. An AI pilot takes 2 to 6 weeks, and scoping sets the timeline of a full build.
Claude Cowork is Anthropic's agent for knowledge work, and it completes tasks across your files and tools, such as building a weekly marketing readout, triaging support tickets or reconciling accounts. Each team adds a plugin with its own skills and connectors, and Anthropic publishes plugins for marketing, sales, customer support, engineering, operations, finance and legal, among others.
Cowork comes with the paid Claude plans: Pro, Max, Ultra, Team and Enterprise, according to Anthropic's Help Center. On Team and Enterprise plans it is on by default and an owner can turn it off, except in organizations that have enabled HIPAA, where it starts off. Enterprise owners can enable it for specific teams through groups and custom roles.
Claude Code for the code itself, and Cowork for the work around it. Anthropic builds Cowork for non-coding knowledge work and positions Claude Code for software engineering, though both run on the same agentic architecture. The Cowork engineering plugin covers standup summaries, incident response, deploy checklists and postmortems, and plugins also work in Claude Code, so a team's skills carry across.
No. Anthropic's Help Center says Cowork sessions cannot be shared with others, although you can share the individual artifacts a session creates. A team shares its setup instead, above all the plugin from the organization's private marketplace, with the skills and the connectors each member's sessions draw on.
Not by default on its commercial products. Anthropic's Privacy Center says it does not use inputs or outputs from commercial products, which it lists as Claude for Work, the Anthropic API and Claude Gov, to train its models unless you send feedback. Local Cowork session data on members' computers falls outside Anthropic's standard retention policies, and admins cannot delete it centrally.
An AI pilot takes 2 to 6 weeks, and scoping sets the timeline of a full build, as Ovidius gives them on its Claude Cowork page. Each rollout tests its connectors in staging on test data, calibrates the approval rules with the people who will use them, and trains every team on its workflow before launch.
Rolling out Claude Cowork team by team?
Book a discovery call. A senior partner runs the call and tells you which of your workflows suit Cowork before anything is scoped.