How to Use Claude Chat, Cowork, and Projects

TL;DR

Use Claude Chat for conversations and Cowork when you want the system to complete a task or produce an artifact. For better results, enable chat search and memory, add persistent instructions, organize ongoing work in Projects, connect approved third-party tools, and save recurring processes as Skills that can be invoked again.

Transcript

After receiving a massive number of updates over the past few weeks, Claude is now more powerful than ever before. But opening it up for the first time may feel confusing if you're relying on any Claude tutorial that's more than a week old. So this video will be the most up-to-date guide you need to get started with the new Claude. I'll cover every... Read More

Key Insights

• Claude's initial setup includes enabling search and reference chats, plus generated memory when that option is available. Memory allows future responses to become more personalized over time, while an import process can bring existing memory from another AI provider into the account.
• Persistent instructions are account-level context that Claude can consider across chats. They can describe who the user is, what Claude will be used for, the desired response style, and other relevant preferences, reducing the need to repeat the same guidance in every conversation.
• Chat is designed for situations where conversation itself is the desired outcome. It supports an ongoing exchange in which a user asks a question, receives a response, and continues with follow-up questions, such as discussing ways to optimize sleep.
• Cowork is designed for tasks where Claude should perform work and create an output. The demonstrated example asks it to research current evidence and anecdotal reports about sleep, then turn the gathered information into an HTML presentation rather than merely providing a conversational answer.
• Projects are dedicated workspaces for organizing chats, Cowork tasks, files, instructions, contacts, and memory around an ongoing topic. Project-specific instructions apply to work started within that project, while project memory develops from the work performed there.
• Chats and Cowork tasks share one dashboard and can coexist inside the same Project. Existing items can be renamed, pinned, or moved into a Project, while frequently used Projects can also be pinned to the sidebar for easier access later.
• Connectors link Claude to third-party tools and their information, and some may permit Claude to write changes back to those tools. The example connects Gmail and Notion so Cowork can find lead updates in email and record them in a Notion sales database.
• Skills are reusable workflows created from tasks a user expects to repeat. After completing a process, the user can ask Claude to turn it into a Skill, save it, and later invoke that workflow instead of reconstructing the original prompt each time.
• More videos from this channel:
  • How to Build a Claude Cowork Automation System
  • How to Create Vertical Shorts With NotebookLM
  • How to start a one person AI business with NotebookLM and Canva
• More from this channel: Paul J Lipsky

Questions & Answers

Q: How should beginners configure Claude before using it?

Beginners should open Settings and review the available capabilities before starting substantial work. The guide recommends enabling search and reference chats and, when shown, generated memory from chat history. Users can also import memory from another AI provider. Under General, they can add persistent instructions covering who they are, their intended uses, preferred response style, and other useful context.

Q: When should you use Claude Chat instead of Cowork?

Claude Chat is appropriate when the conversation itself is the goal, especially for questions, explanations, brainstorming, and follow-up discussion. The guide illustrates this with a conversation about optimizing sleep. Cowork is more suitable when Claude needs to carry out a task and deliver an artifact, such as researching sleep and creating an HTML presentation from the results.

Q: What is Claude Cowork used for?

Claude Cowork is used to delegate work rather than only exchange messages. A user can describe a goal, ask Claude to gather relevant information, and request a finished output. In the example, Cowork researches current sleep evidence and anecdotal reports, then builds a presentation. It can also work through connected services to inspect information and update external records.

Q: How do Claude Projects organize ongoing work?

Claude Projects provide dedicated workspaces for an ongoing task or topic. Each Project can contain chats, Cowork tasks, files, contacts, instructions, and shared project memory. Project instructions apply to new work started there, while memory becomes more helpful as Claude learns about the work. Existing chats and tasks can also be moved into a Project after they were created.

Q: How do you create and configure a Claude Project?

To create a Project, open Projects, choose New Project, provide a name, and create it. Within the Project, use the plus controls to add project-specific instructions and contextual files. The demonstrated sources include files from a device, GitHub, or Google Drive. Frequently used Projects can be pinned so they remain readily accessible in the left sidebar.

Q: What are Claude Connectors and how do they work?

Connectors link Claude with third-party tools so it can access information stored in those services and, depending on the connection, make changes there. Users can manage existing connections or browse for additional ones. The guide names Gmail, Google Drive, Calendar, and Notion, then demonstrates Claude reading Gmail updates and recording relevant activity in a Notion sales database.

Q: How can Claude update a sales CRM using Gmail and Notion?

With Gmail and Notion connected, a user can ask Cowork to inspect the inbox for new messages involving leads already listed in a Notion sales CRM. Claude can log the relevant email activity in those lead records, flag the hottest deals with comments, and return a report describing what it found. This task is assigned to Cowork because it requires operational work.

Q: How do you turn a repeated Claude workflow into a Skill?

After Claude successfully completes a workflow, ask it to convert that process into a Skill and describe when the Skill should run. In the example, the saved workflow checks Gmail for updates from leads, updates the Notion CRM, and reports the results whenever the user asks to update the lead CRM. Saving it avoids repeating the complete original prompt later.

Summary & Key Takeaways

• Claude begins with a familiar prompt box, but its effectiveness depends on a few initial settings. Users can enable chat search, references, and memory, import memories from another AI provider, and add persistent instructions describing their identity, intended use, preferred response style, and any other context Claude should retain across conversations.

• The Home interface supports Chat and Cowork alongside a separate Claude Code area for developers. Chat suits back-and-forth conversations, while Cowork is intended for delegated work such as researching a topic and producing a presentation. Both modes share the dashboard, settings, instructions, Projects, Connectors, and Skills.

• Projects organize related chats, Cowork tasks, files, instructions, contacts, and project memory within dedicated workspaces. Connectors provide access to third-party tools such as Gmail, Google Drive, Calendar, and Notion. A successful recurring process, such as synchronizing email updates with a Notion sales database, can be saved as a reusable Skill.