What Is Claude Cowork and How Does It Work?

TL;DR

Claude Cowork is Anthropic's native desktop app that, unlike Claude Chat, accesses your local files and delivers ready-to-use output directly into your folders instead of a chat window. It removes Chat's 20-file, 30MB-per-file limits, offers a larger context window, and stores persistent memory in files so it remembers preferences across sessions.

Transcript

You're in the right place if you use claw chat every day, are intimidated by cloud code, and know co-work exists, but not why you need it. Because in this video, we're breaking down the seven core capabilities of cloud co-work, each with realworld use cases, so that by the end, you know exactly what it can do and where it fits in your workflow. Qui... Read More

Key Insights

• Claude Cowork differs from Claude Chat in three core ways: it accesses local files (removing Chat's 20-file-per-conversation and 30MB-per-file limits), it delivers finished files to your folder instead of a copyable chat response, and it needs outcome-first prompting.
• Claude Chat uploads files to the cloud, capping you at 20 files per conversation and 30 megabytes per file, while Cowork's local file access removes both limits and handles more files at larger sizes.
• Cowork has a larger context window than Chat, letting you do more work before triggering the compacting function, where Claude summarizes prior context to free space but may lose important details.
• Prompting Cowork uses outcome-first language: you define the end result, constraints, and quality bar, then let it run, rather than the task-first step-by-step instructions Chat requires.
• Recommended setup includes selecting the Cowork tab, adding Cowork-only instructions as guardrails, enabling both memory features under Capabilities, setting tool access to load-when-needed, and creating a dedicated 'cowork playground' subfolder to contain all work.
• Cowork is strict about file access: a file dragged in from another folder like Downloads can be loaded but not read, so files must live inside the folder Cowork was granted access to.
• Capability one is local file access: Cowork can create, edit, and organize files directly on your computer, such as reading 100+ PDF and JPEG receipts and outputting a formatted Excel expense report with flagged rows to double-check.
• Persistent memory works because Cowork saves memory to actual files (like CLAUDE.md and memory.md) on your computer, so it can hold every decision and preference for as long as needed, unlike Chat's online memory which has a hard storage limit.
• Connectors let Cowork reach into tools you already use, like Gmail, Google Calendar, Google Drive, and Notion, so it can read from and work directly inside them, for example extracting your writing tone from a month of emails.

Questions & Answers

Q: What is the difference between Claude Chat and Claude Cowork?

Claude Chat is the familiar cloud chatbot, while Claude Cowork is Anthropic's native desktop app. Three core differences: Chat uploads files to the cloud with a 20-file-per-conversation and 30MB-per-file limit, while Cowork accesses local files without those limits. Chat returns a response you must copy or download, while Cowork delivers ready-to-use files directly in your folder. Cowork also has a larger context window, letting you work longer before compacting.

Q: What are the file upload limits in Claude Chat?

Claude Chat uploads files to the cloud, which limits you to 20 files per conversation and 30 megabytes per file. These limits exist because the files must be sent to the cloud. Claude Cowork removes both restrictions because it accesses your local files directly, so it can easily handle more files at larger sizes without needing to upload them anywhere.

Q: How should you prompt Claude Cowork?

With Cowork, you use outcome-first language, where you define the end result, the constraints, and the quality bar, then let Cowork run and complete the work. This contrasts with Claude Chat, which uses task-first language where you tell it what to do step by step and then do the work yourself. For example, you tell Cowork you have 15 raw thumbnail photos that need organizing into subfolders by topic with descriptive names, and it completes the task.

Q: How do you set up Claude Cowork properly?

Select the Cowork tab, then under settings use the Cowork tab to add instructions that apply to Cowork only, such as guardrails that stop it from touching files without asking. Under Capabilities, enable both memory features, set tool access to load tools when needed, and enable the listed features. Finally, create a subfolder in your documents called 'cowork playground' so all work stays contained and other files remain untouched.

Q: Why can't Cowork read a file I dragged in from another folder?

Cowork is very strict about file access. If you drag a file from your Downloads folder into Cowork, it will be able to load the file but it cannot read it. To work with files, they must be within the folder you granted Cowork access to, such as the cowork playground folder. When you first give access to a folder, Cowork asks for permission and you can click 'always allow' for that folder.

Q: What is Capability #1, local file access, and what can it do?

Local file access means Cowork can create, edit, and organize files directly on your computer. Examples include reading over 100 receipt images (PDFs and JPEGs) to extract details and output a formatted Excel expense report with date, vendor, category, amount, and totals, flagging unclear rows. It can also split a large 400+ megabyte PDF into per-chapter files, or rebuild non-editable Notebook LM slides into an editable PowerPoint with real text boxes.

Q: How does persistent memory work in Claude Cowork?

Persistent memory means you tell the AI something and it remembers that in future conversations without you repeating yourself. Because Claude Chat's memory is stored online, it has a hard limit. Cowork instead saves memory to actual files on your computer, so it can remember every decision and preference for as long as you need. When you tell Cowork to remember something, it writes to files like CLAUDE.md and memory.md, and the more it writes, the better it works your way.

Q: What are connectors in Claude Cowork and how do you set them up?

By default Cowork only sees what's in your folder, but connectors let it reach into tools you already use like Gmail, Google Drive, and Notion, so it can read from them and work inside them. To set up, click the Customize tab on the left, go to Connectors, click the plus icon, browse connectors, and search for your tool. For example, once Gmail is connected, you can have Cowork read a month of emails to extract and save your tone of voice as writing style principles.

Summary & Key Takeaways

• Claude Cowork is Anthropic's native desktop app that turns Claude from a chatbot into a productivity system. The video breaks down seven core capabilities with real-world use cases so viewers understand what Cowork does and where it fits in their workflow versus Claude Chat.

• Three key differences separate Chat from Cowork: Cowork accesses local files (removing the 20-file and 30MB limits), delivers finished files to your folder rather than a chat window, and uses outcome-first prompting where you define the result, constraints, and quality bar.

• Local file access powers persistent memory: Cowork writes preferences and decisions to CLAUDE.md and memory.md files, remembering them across brand-new sessions. Connectors extend it into Gmail, Drive, and Notion, letting it read emails to learn your tone or compare Gemini transcripts against Notion notes.