Claude Cowork adds CLAUDE.md workflows

- Anthropic has expanded Claude Cowork with reusable workflow patterns built around CLAUDE.md-style instructions, pushing its desktop agent beyond one-off chat tasks. - Anthropic’s materials show Cowork handling scheduled reports, file cleanup, spreadsheet extraction, and document drafting across local files and connected apps. - The shift fits Anthropic’s broader push to package repeatable agent behavior for teams, not just individuals. (anthropic.com)

Anthropic is turning Claude Cowork into a desktop agent that follows repeatable workflows, not just a chat window that answers one prompt at a time. (anthropic.com) On Anthropic’s Cowork product page, the company says Claude can work across local files, folders, and everyday applications to return a finished deliverable. It lists organizing files, preparing documents, synthesizing research, and extracting data from dense records as core use cases. (anthropic.com) A second Cowork page shows the product running scheduled tasks, including pulling metrics into a weekly report every Friday, checking email each morning, and generating recurring Slack digests. Anthropic also shows Cowork turning receipts, invoices, and screenshots into formatted spreadsheets. (claude.com) That matters because Anthropic is selling a different model of office AI. Instead of asking workers to break a project into many prompts, Cowork is pitched as software that takes the outcome, chooses steps, and moves between tools on its own. (anthropic.com) (claude.com) The workflow piece comes from Anthropic’s broader documentation around reusable instructions. In a guide published through Anthropic resources, the company defines a “skill” as a folder that packages instructions, scripts, references, and assets so Claude can perform a task the same way each time. (resources.anthropic.com) That guide says skills are meant for repeatable work such as research methods, document generation, style-guide compliance, and multi-step processes. Anthropic says teams can use them to standardize how Claude behaves across an organization. (resources.anthropic.com) (anthropic.com) Anthropic has described the same pattern more explicitly on the coding side. A March 24, 2026 webinar deck on advanced Claude Code patterns says teams can “control Claude’s behavior with CLAUDE.md and Hooks” and orchestrate subagents for parallel work. (resources.anthropic.com) Anthropic’s research and policy writing now treats Cowork and Code as part of the same agent family. In an April 9, 2026 paper on trustworthy agents, the company says products like Claude Code and Claude Cowork can write code, manage files, and complete tasks across multiple applications. (anthropic.com) Cowork is also moving from experiment to rollout. On April 9, 2026, Anthropic added enterprise controls including role-based access, group spend limits, usage analytics, OpenTelemetry support, a Zoom connector, and per-tool connector controls; 9to5Mac reported Cowork was then generally available for paid subscribers on macOS and Windows. (9to5mac.com) The throughline is that Anthropic is packaging repeatable agent behavior as a product feature. Cowork is no longer just “Claude on your desktop”; it is being positioned as a way for teams to hand off recurring document, spreadsheet, and file work to a system that follows saved instructions. (anthropic.com) (resources.anthropic.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.