💡 Just started using Claude Cowork? Don't touch anything else until you've done these five things. This is the difference between a generic AI and one that actually works the way you do.
📋 Overview
Claude Cowork is most powerful when it's configured around you — your workflows, your output preferences, and the tools you actually use. This guide walks through the exact setup I run on every fresh account.
| ⚡ Step | What You're Doing | Where |
| 1️⃣ | Let Claude interview you & build your instructions | Settings |
| 2️⃣ | Enable memory toggles | Capabilities |
| 3️⃣ | Enable artifact & app generation | Capabilities |
| 4️⃣ | Install the right plugins | Customize |
| 5️⃣ | Install the right skills | Customize |
⏱️ Total setup time: ~10 minutes
🛠️ Step-by-Step Setup
1️⃣ Have Claude Interview You
The fastest way to get a tailored Claude is to let it ask you the questions. This generates a global instructions file (your CLAUDE.md) that informs every future conversation.
<details><summary>📝 Copy-paste interview prompt (upgraded)</summary>
Interview me one question at a time to build the best possible
CLAUDE.md file for global instructions.
Ask me about:
- My role, industry, and primary goals
- The types of projects and tasks I'll bring to you
- My preferred tone, format, and level of detail in outputs
- Tools and platforms I work in daily
- Common mistakes you should avoid when working with me
- How I like to receive feedback and revisions
Ask follow-up questions where my answers are vague. Once you have
enough, generate a clean, comprehensive CLAUDE.md file I can paste
directly into my global instructions.</details>
Once Claude generates the document:
- 📋 Copy the full output
- ⚙️ Open Settings in Cowork
- 📄 Paste it into the instructions field
- 💾 Click Save
💡 Treat your CLAUDE.md as a living document. Update it any time your workflow or preferences shift.
2️⃣ Turn On Memory 🧠
Inside Capabilities, flip on these two toggles so Claude never forgets who you are or what you've discussed.
| 🔘 Toggle | What It Does |
| 🧠 Memory | Remembers your preferences & context across sessions |
| 💬 Conversation History | Recalls past conversations for continuity |
💡 With these on, you stop re-explaining yourself every single chat. Massive time-saver.
3️⃣ Enable Artifacts & Apps 🎨
Still inside Capabilities, turn on these three toggles so Claude can build for you, not just talk to you.
| 🔘 Toggle | What You Unlock |
| 📄 Artifacts | Documents, code, and structured outputs |
| 🤖 AI-Powered Apps | Interactive mini-apps built on the fly |
| 📊 Visualizations | Charts, diagrams, and visual data |
💡 This is where Claude goes from "chatbot" to "actually builds things." Don't skip it.
4️⃣ Install the Right Plugins 🔌
Head into Customize to plug Claude into the tools and capabilities you need.
How to install:
- ➕ Click the plus button
- 🗂️ Select Browse Plugins to open the full directory
- ✅ Choose anything you want
- ➕ Click the plus button on each to install
<details><summary>💡 Pro tip: How to choose plugins</summary>
Don't install everything just because you can. Start with plugins that map to your most frequent tasks — research, writing, data, or whatever your CLAUDE.md says you do most. You can always add more later.
</details>
5️⃣ Install the Right Skills 🧩
Skills give Claude specialized abilities for specific kinds of work.
How to install:
- 🗂️ Open the directory and click Skills
- ✅ Select all the ones you want
- ➕ Click the plus button to install
| 🔌 Plugins vs 🧩 Skills | Plugins | Skills |
| Purpose | Connect tools & integrations | Add specialized abilities |
| Best for | Extending what Claude can access | Improving how Claude performs tasks |
| Where to find | Browse Plugins directory | Skills tab in directory |
📚 Prompt Library
<details><summary>🔧 Build your CLAUDE.md (full interview prompt)</summary>
Interview me one question at a time to build the best possible
CLAUDE.md file for global instructions.
Ask me about:
- My role, industry, and primary goals
- The types of projects and tasks I'll bring to you
- My preferred tone, format, and level of detail in outputs
- Tools and platforms I work in daily
- Common mistakes you should avoid when working with me
- How I like to receive feedback and revisions
Ask follow-up questions where my answers are vague. Once you have
enough, generate a clean, comprehensive CLAUDE.md file I can paste
directly into my global instructions.</details>
<details><summary>♻️ Refresh your instructions later</summary>
Here's my current CLAUDE.md file: [paste it].
Ask me what's changed in my workflow, role, or output preferences
since this was written. Then rewrite the file to reflect my updated
needs while keeping anything that's still accurate.</details>
✅ Final Setup Checklist
| Done? | Task |
| ⬜ | Ran the interview prompt & saved CLAUDE.md in Settings |
| ⬜ | Enabled Memory + Conversation History |
| ⬜ | Enabled Artifacts, AI Apps, & Visualizations |
| ⬜ | Installed relevant plugins |
| ⬜ | Installed relevant skills |
⭐ Pro Tips
<details><summary>🎯 Start lean, then expand</summary>
Resist installing every plugin and skill on day one. Set your foundation, use Claude for a week, and add tools only when you hit a real need. A focused setup beats a bloated one.
</details>
<details><summary>🔁 Revisit your config monthly</summary>
Your work evolves — your setup should too. Re-run the refresh prompt and audit your plugins/skills regularly so Claude keeps pace with you.
</details>
💬 Want More?
- 🛠️ Want me to personally set up your Claude? Comment "set up"
- 📖 Want the full optimization guide? Comment "Claude" and I'll send it over
- ➡️ Follow to stay ahead in the world of AI