Getting Started with Claude Cowork: A Simple Setup Guide
An assistant that has to be told where every file lives, every time, is not much of an assistant. Claude Cowork can read your files and tools, but only after you connect them once. Setup is the difference between a chatbot and a coworker.
Create a project around a folder
Open the Claude desktop app and click Cowork, then create a Project and point it at an existing folder. The folder matters: Claude reads every file inside it, like PDFs, spreadsheets, and Word docs, and saves finished work back to the same place. Point it at your Work Reports folder and it learns the format of old reports and matches them.
Connect your tools once
In the Customize sidebar, connect Slack, Google Drive, Calendar, Gmail, and whatever else holds your work. After that, "check what the team said in Slack about this week's launch" is a real instruction: Claude finds the messages, drafts the Gmail reply, saves the file to Drive.
Write down how you work
Inside each Project there is an Instructions box. Put everything you repeat: your role, output formats (always save as .docx), where things live, judgment calls to make the same way every time. Global Instructions in Settings apply across all projects, like ask before deleting anything.
Extras worth knowing
Plugins add domain expertise, pre-built for finance, sales, legal, and more. Scheduled tasks: type /schedule, describe what to run and when, and the weekly report becomes a file waiting on Friday morning while the desktop app is open. The Chrome extension lets Cowork work inside sites you are logged into; only enable it where you trust the site.
How a task runs
Describe what you want and where the result should go. Claude reads your files, asks a few clarifying questions, then works in the background with a progress panel showing each step. The result lands where you pointed it.
The trade-off
Cowork does not remember between sessions: anything it should know every time belongs in Project instructions, global instructions, or a file in the folder. And big tasks spend real usage: reading many files and running long jobs adds up, so check Settings then Usage to see where you stand. Plan requirements and limits change; check Anthropic's current docs for what Cowork needs.
Your first task
Try this: point a Project at a real folder and ask for a one-page summary of everything related to one project, matching the format of an existing file.
Recommended for you
- ContextClaude Code
The MCP You Connected Is Invisible to Your Spawned Agent
MCPs you wire into your main session do not cross into spawned sub-agents. The agent reports Linear unavailable and invents its scope from prompt text.
- ContextClaude Code
Stop Describing Your Codebase. Show It to the Agent.
You described your project in English. The agent produced generic code that matches none of it. Stop describing. Mention the file. Run the command.
- ContextClaude Code
Why Your AI Agent Ignores Your Team's Conventions
You typed the rule in chat three times and the agent still broke it. Chat is for the current task. Project rules live in AGENTS.md. Yours is empty.
- ContextAmpClaude Code
Edit, Restore, or Handoff: The Three Reset Moves Every Agent User Needs
Three things break an agent session: a wrong turn, a drifted conversation, a thread too big. Three moves fix each: Edit, Restore, Handoff. Picking the wrong one wastes time.
Enjoyed this article?
Subscribe for new articles. No spam. Unsubscribe anytime.