Video Summary
Claude Cowork is a native desktop app that accesses local files, avoids chat file-size limits, and writes outputs directly to folders.
Use outcome-first prompts for Cowork (define the result, constraints, and quality bar) rather than step-by-step task prompts.
Enable persistent memory so Cowork saves preferences and corrections to local files for ongoing, personalized automation.
Connectors (Gmail, Google Drive, Notion, etc.) let Cowork read external tools and produce polished outputs in those platforms.
Skills are reusable, testable workflows you build in Cowork to automate repeated tasks like text refinement or weekly reports.
Cowork is a native desktop app that can access and modify local files, supports a larger context window, and writes ready-to-use outputs directly into folders, whereas Chat uploads files to the cloud and returns responses in a chat window.
Local file access lets Cowork read, create, edit, split, and organize large numbers of files (including big PDFs and many receipts) without the 20-file or 30 MB-per-file limits of Chat.
Persistent memory saves preferences, corrections, and rules to files on your computer so Cowork remembers past decisions and improves automation across sessions.
Connectors let Cowork access external apps like Gmail, Google Drive, and Notion so it can read data, analyze tone, cross-reference notes/transcripts, and produce polished pages in those platforms.
Skills are reusable, task-oriented workflows you create and test inside Cowork (e.g., text refinement or report assembly). Build the first few yourself to learn the process before relying on templates.
The browser extension is currently slow and unreliable for some workflows; scheduled tasks and local automations are more dependable, and skills/memory don’t auto-transfer between machines unless backed up.
"Claude Chat uploads files to the cloud, while Co-work can access your local files, allowing it to handle more files with larger sizes."
Claude Chat operates within strict file limitations, allowing only 20 files per conversation and a maximum of 30 megabytes per file, which can hinder productivity.
In contrast, Claude Co-work avoids these restrictions by accessing local files, enabling it to manage a greater number of files and larger file sizes without issues.
Additionally, Co-work possesses a broader context window, allowing users to maintain more extensive ongoing work before necessitating a summary of prior contexts.
"Co-work delivers ready-to-use files directly in your folder, while Chat requires users to format and download responses themselves."
With Claude Chat, users receive responses in a chat window, but they must manually format, copy, and download the output.
Conversely, Claude Co-work streamlines this process, autonomously performing tasks and providing polished outputs directly into specified folders, thus improving efficiency.
"With Chat, we use task-first language, while with Co-work, we use outcome-first language."
Users engage with Claude Chat by detailing sequential tasks, such as asking it to review files and recommend organization methods.
In contrast, with Co-work, users define the desired outcome and parameters, delegating the execution entirely to the tool. This approach allows Co-work to function autonomously, completing tasks efficiently, such as sorting and naming files.
"These instructions act as guardrails that prevent Co-work from altering your files without consent."
Before using Co-work, users should adjust settings to ensure its features are optimized for personal preferences and workflows.
Personal preferences can be set under the general settings tab, while specific instructions for Co-work can be added to its dedicated settings tab.
Utilization of a starter pack with pre-prepared instructions enhances the user experience by providing guidelines that restrict file modifications, such as requiring confirmation before deleting or renaming any files.
"Co-work can create, edit, and organize files directly on your computer."
One of the key functionalities of Claude Co-work is its ability to manipulate files locally. Users can direct Co-work to generate tasks such as creating expense reports from receipts by specifying details and criteria relevant to their needs.
This capability allows Co-work to process a larger volume of documents compared to Claude Chat, which is limited by its file handling parameters.
Furthermore, Co-work can segment large PDFs into manageable sections, making them easier to work with on other platforms or tools.
"Co-work saves memory to actual files on your computer, meaning it can remember every decision you've asked it to hold."
Co-work's persistent memory function significantly enhances user experience by retaining context and preferences over time, unlike Claude Chat, which has limitations due to online memory constraints.
For example, users can query Co-work for historical data, such as past newsletters produced, without needing to reintroduce previous discussions.
This functionality extends to adjusting generated outputs based on user feedback, ensuring personalized interaction that builds upon user input over subsequent sessions.
"By default, Co-work can only see what's in your folder."
Claude Cowork allows users to connect external tools such as Gmail, Google Drive, and Notion by utilizing connectors. This enables Claude to access these applications directly, extracting information for seamless integration.
To set up these connections, navigate to the customize tab, select connectors, click the plus icon, browse for your preferred tools, and make the connections.
It is recommended to connect essential tools like Gmail, Google Calendar, Google Drive, and Notion to maximize the functionality of Claude Cowork.
"Co-work then proceeds to read through my emails, locates those with substantive writing, distills the patterns it sees, and presents them to me."
Once Gmail is connected, users can instruct Co-work to analyze past emails to determine their tone of voice. This analysis results in saved writing style principles that guide future correspondence.
For instance, a user might instruct Co-work to extract their tone of voice from the past month of emails, allowing it to better understand and replicate the user's communication style.
"Co-work will read both documents, compare what was said versus what was written, and surface the action items."
Co-work can help manage meeting notes by cross-referencing transcripts from tools like Google Gemini and notes in Notion. It identifies commitments or action items that may not have been captured during the note-taking process.
This capability enhances productivity by ensuring that no important details are overlooked, making it easier to track tasks and responsibilities.
"Co-work gives me a version two of the output."
Users can develop skills in Co-work, which are task-oriented functions that streamline workflows. For instance, if a user wants to refine text output for clarity, they can create a skill that systematically revises text according to their specified criteria.
Skills can be created by providing Co-work with instructions, and after testing, they can be easily added to the user’s repertoire. Future requests for text refinement will now utilize the newly created skill, offering efficiency in the process.
The process of skill creation highlights the adaptability of Co-work, as users can not only make simple skills but also more complex workflows through repeated steps.
"So I share the raw updates from the three teams with Co-work and say combine these three team updates into one formatted weekly marketing update."
A practical application of skills involves generating comprehensive weekly updates from disparate team reports. Users can provide raw content and instructions on desired format and structure, allowing Co-work to compile and refine the information efficiently.
Once a draft is prepared, users can provide feedback for further refinements, showcasing the ability of Co-work to iterate based on user input until the final output meets their expectations.
"While there are websites that offer skill templates, I highly recommend creating the first few yourself."
To effectively create skills, it is essential to understand Co-work's functionality firsthand. Users should start by building skills from scratch to gain familiarity with the process.
Skills can be updated and overwritten, ensuring continuous improvement. It is advised to back up skills externally, as they do not transfer automatically to new computers.
Users are encouraged to learn the workflow and create skills afterward for a more efficient process rather than relying on predefined templates.
"Co-work projects can write to their knowledge files directly, unlike Claude chat projects."
Although both Co-work and Claude Chat manage projects, Co-work enhances productivity through features such as local file access, persistent memory, connectors, and skills.
Co-work projects allow users to codify learnings directly into their knowledge files, which can streamline processes and ensure that new information is preserved without duplicate actions required.
Understanding these distinctions can guide users in selecting the right Claude project for their tasks based on their specific needs.
"Co-work can theoretically hand off tasks to the browser extension."
The integration of a browser extension within Co-work is currently hindered by issues like slow performance, unreliability, and excessive usage consumption during interactions.
Despite these limitations, Co-work excels in scheduling tasks, as it executes them reliably, specifically when set up for tasks such as daily inbox triage.
Utilizing scheduled tasks can streamline daily responsibilities effectively if users adequately set up these functionalities within the Co-work framework.
"The drafted replies are basically flawless."
Claude Cowork utilizes an inbox triage workflow that streamlines email management by following a set of prescribed rules. This workflow was influenced by a teaching methodology from the Workspace Academy course.
The first capability, local file access, enables Claude Cowork to save these rules in a markdown document, creating a comprehensive set of instructions for the user.
Through the connectors feature, which links Claude Cowork to Gmail, the application can read incoming emails and match them against the user’s inbox zero rules, allowing it to generate contextually relevant email replies.
"Thanks to persistent memory, Claude Cowork remembers those corrections and now triages my inbox the way I would."
Initially, the system did not function optimally and required user feedback to improve its responsiveness to specific email phrasing. The user provided corrections, helping fine-tune how Claude Cowork would draft replies.
The persistent memory feature is crucial, as it stores feedback over time, allowing Claude Cowork to accurately mimic the user’s triaging preferences based on past interactions.
Within the application’s structure, specific subfolders, such as the “email HQ” subfolder, store the inbox zero triage workflow and memory documentation, ensuring that the system can recall user preferences effectively.
