Learn the Core Features of Claude Co-work
If you use Claude chat daily, feel a bit lost with Claude Co-work, and want to understand why you need it, this guide is for you. We’ll break down the main features of Claude Co-work. You’ll learn what it can do and how it fits into your daily tasks.
Understanding the Differences: Claude Chat vs. Claude Co-work
Claude Co-work is a desktop app, and it’s different from the Claude chat you might be used to. Here are the key differences:
1. File Handling and Context Window
Both Claude chat and Co-work need internet. However, Claude chat sends files to the cloud. This means you have limits: only 20 files per chat and 30 megabytes per file. Claude Co-work, on the other hand, can access files directly on your computer. This removes those file limits. It can handle many more files, even very large ones. Co-work also has a much bigger ‘context window’. This is like its short-term memory. A larger window means you can do more work before Claude needs to summarize your conversation, which can sometimes lose details.
2. Output and Integration
With Claude chat, you get responses in a chat window. You then have to copy, format, or download these results yourself. Claude Co-work does the work for you. It can create ready-to-use files and put them right into your computer folders. It can also connect to other apps like Notion to create polished pages there.
3. Prompting Style
Because of these differences, you prompt Co-work differently than chat. With Claude chat, you often use ‘task first’ language. You tell it exactly what to do, step-by-step. For example: ‘Review my photos and suggest a naming system.’ Claude chat gives you a text suggestion, and you do the rest. With Co-work, you use ‘outcome first’ language. You describe the final result you want, including any rules or quality standards. For example: ‘I have 15 photos in this folder. Organize them by topic with good file names.’ You let Co-work handle it, and the work is done quickly.
Setting Up Claude Co-work
Let’s get Co-work set up correctly. First, find the ‘Settings’ menu. Under the ‘General’ tab, you’ll see personal settings for both chat and Co-work. Go to the ‘Co-work’ tab to add instructions that only apply to Co-work. These act like safety rails. For instance, you can tell Co-work to always show you what changes it plans to make before deleting, overwriting, or renaming any files. It will wait for your okay. Think of these as training wheels – you can remove them as you get more comfortable.
Enable Key Features
In ‘Settings’ under ‘Capabilities’, make sure to enable ‘memory features’. You can turn off ‘location metadata’ for better privacy. Set ‘tool access’ to ‘load tools when needed’ and enable other features listed there.
Create a Dedicated Folder
On your computer, create a new folder named ‘Co-work Playground’ in your Documents folder. This keeps all your Co-work activities in one place, protecting your other files.
Getting Started with Your First Conversation
When you start a new conversation in Co-work, point it to your ‘Co-work Playground’ folder. If it’s the first time, Co-work will ask for permission to access the folder. Click ‘Always allow’. Remember, Co-work is strict about file access. If you drag a file from another folder into Co-work, it can load it, but it can’t read it if it’s not within your allowed playground folder. So, make sure all files you want to work with are inside your ‘Co-work Playground’ folder.
Core Capability 1: Local File Management
Co-work can create, edit, and organize files directly on your computer. It’s like having a smart assistant for your file system.
Use Case: Expense Report from Receipts
Imagine you have over 100 receipt photos (PDFs and JPEGs) in your playground folder. You can ask Co-work to create an expense report. Specify the columns you need: date, vendor, category, and amount. You can also tell it to flag any blurry or unclear entries. Co-work will first show you the steps it plans to take and wait for your approval. Then, it reads each image, pulls out the needed information, and creates a formatted Excel file right in your folder. It will also mark any rows that need your double-check. This is hard to do with Claude chat because of its file limits and the fact that it only gives you text output.
Use Case: Splitting Large PDFs
If you have a huge PDF file that’s too big for other tools, Co-work can help. You can ask it to break the PDF into smaller files, perhaps one for each chapter or section. It can also create descriptive file names so you can easily find what you need. Unlike Claude chat, Co-work can identify natural breaks in the document and create all the separate files for you.
Use Case: Rebuilding Editable Presentations
Sometimes, presentations are made of images, making them hard to edit. You can give such a file to Co-work and ask it to rebuild the presentation with actual text boxes. You’ll get the same content and slide order, but now you can modify the text. While not always perfect, it’s much better than working with static images.
Core Capability 2: Persistent Memory
This is a powerful feature that relies on Co-work’s ability to manage local files. It means Co-work can remember things over long periods, far beyond a single conversation.
How it Works
Claude chat has a form of memory, but it’s stored online and has limits. Co-work saves its memory directly to files on your computer. This means it can remember every preference and decision you’ve made for as long as you need it to. For example, you can ask Co-work how many newsletter editions you’ve created together and ask it to break them down by topic. It can recall this information from past interactions, even in a new session, because it saved that data locally. When you give Co-work feedback or set preferences, it writes this information into special files like ‘memory.md’ or ‘cla.md’ in your playground folder. The more it writes, the better it learns to work the way you prefer.
Use Case: Meeting Summaries and Preferences
You can have Co-work summarize a meeting transcript. If you don’t like the structure of the summary, you can adjust it and tell Co-work to compare its version with yours and save those preferences. Co-work will then reconcile the changes and, importantly, create or update those ‘cla.md’ and ‘memory.md’ files. These files help Co-work remember your specific needs for future tasks.
Core Capability 3: Connectors
By default, Co-work can only access files in your chosen folder. Connectors allow Co-work to interact with other tools you use, like Gmail, Google Drive, or Notion. It can read from them and perform actions directly within them.
Setting Up Connectors
To set up connectors, go to the ‘Customize’ tab on the left, then select ‘Connectors’. Click the plus icon, then ‘Browse connectors’. Search for the tools you use. For basic use, connecting Gmail, Google Calendar, Google Drive, and Notion is recommended.
Use Case: Analyzing Email Tone
After connecting Gmail, you can ask Co-work to understand your writing tone. It can read your emails from the past month, extract your tone, and save these as writing style principles. The next time you draft an email, Co-work can use these principles to help you maintain your specific tone of voice.
Use Case: Cross-Referencing Meeting Notes
Imagine your team takes notes in Notion, but meeting transcripts are generated by Google Gemini and stored in Google Drive. You can ask Co-work to compare the Gemini transcript with the Notion notes and find any commitments that were missed in the written notes. Co-work will read both sources, compare them, and highlight any action items that fell through the cracks. This shows how two connectors can work together through Co-work.
Core Capability 4: Skills
Skills are custom actions you teach Co-work to perform. Once created, you can use them with a single click.
Creating a Skill
Start by asking Co-work to perform a task, like making text more clear and concise. Once you’re happy with the output, you can tell Co-work to turn that specific process into a ‘skill’. Co-work will guide you through creating this skill, asking clarifying questions about what it should do and how it should be applied. After creation, you can add it to your Co-work. When you give Co-work a new piece of text, you can simply tell it to run your ‘clear and concise’ skill on it, and it will repeat the process automatically.
Use Case: Consolidating Team Updates
If you receive weekly updates from different teams in various formats, you can teach Co-work how to combine them into one clean report. You provide the raw updates and specify the desired format (e.g., three key metrics, three highlights, three lowlights, under 300 words). After Co-work generates a draft, you give feedback. Once you’re satisfied, you can tell Co-work to create a ‘weekly report’ skill based on this entire workflow. This skill can then be used repeatedly for future updates, saving significant time.
Managing Skills
To create your own skills, ensure the ‘Skill Creator by Anthropic’ skill is enabled in ‘Customize’ > ‘Skills’. It’s best to create your first few skills yourself to understand the process. You can update existing skills by telling Co-work which skill to change and how. After it recreates the skill, you need to save the new version. Remember to back up your skills, as they don’t automatically transfer to a new computer.
Core Capability 5: Projects
Co-work projects are similar to Claude chat projects but include all the advanced capabilities like local file access, persistent memory, connectors, and skills.
Writing to Knowledge Files
A key difference is that Co-work projects can write to their knowledge files directly. For example, if you have a project that helps make your writing more articulate, and Claude learns a new principle about your style (like starting explanations with an example), Co-work can directly update the project’s instruction file with this new rule. In contrast, with Claude chat, you might have to manually update or replace old knowledge files.
Core Capability 6: Browser Extension (with Caveats)
If you have the Claude browser extension installed, Co-work can theoretically hand off tasks to it. However, the extension is currently slow, unreliable, and uses a lot of your usage allowance because it overthinks every step. The speaker does not recommend using it at this time. This also means that when Co-work needs to search the web, it might fall back on this less reliable extension, which is a limitation compared to Claude chat where you have more control over web searches.
Core Capability 7: Scheduled Tasks
While other AI tools have scheduled tasks, Co-work’s implementation works very well, thanks to its other capabilities.
Use Case: Inbox Triage
You can set up a scheduled task, like an ‘inbox triage’ that runs every day at 6 a.m. This task can produce a report and draft replies for your emails. Co-work uses your local file access to store your inbox workflow rules. Using connectors, it connects to Gmail to read your emails. It then maps these emails against your rules and its understanding of you to draft relevant replies. Because of persistent memory, Co-work learns from your feedback over time, refining its ability to triage your inbox exactly how you would. This involves setting up specific folders and ensuring Co-work has access to your local memory files where it stores your preferences and feedback.
Conclusion
Claude Co-work offers powerful capabilities for managing files, remembering information, connecting to tools, and automating tasks. By understanding and utilizing these core features, you can significantly enhance your workflow.
Source: Learn 80% of Claude Cowork in Under 20 Minutes (YouTube)
