Most people open Claude, type a question, read the answer, and close the tab.
Save this :)
They are using maybe five percent of what is sitting right in front of them.
A small group of people have actually set Claude up. They have it remembering their preferences, running on a schedule, connected to their files, reviewing its own work, and building tools that didn't exist an hour ago. To them, Claude is not a chat box. It is an operating system for their work.
The difference is not a secret prompt. It is setup.
Below are 50 features most users have never touched. Work through them and you will get more out of Claude in a weekend than most people get in a year. I've grouped them so you can build your setup in layers.
Layer 1: The Foundation Settings (Set These First)
1. User Preferences. Tucked in settings, this is a standing instruction Claude applies to every conversation. Drop your tone, your formatting rules, your defaults here once and stop repeating yourself forever.
2. Custom Styles. Separate from preferences, styles let you save and switch between writing voices. Build one for client emails, one for X posts, one for technical docs. Switch with a tap.
3. Model Switching Mid-Conversation. You are not locked into one model. Use the most powerful option for hard reasoning, a faster one for quick tasks, and switch without losing your thread.
4. The Web Search Toggle. Off by default in many setups. Flip it on and Claude pulls live information and cites its sources inline instead of relying only on what it already knows.
5. Incognito Conversations. A chat that leaves no trace and never feeds memory. Use it for anything sensitive or anything you don't want shaping future responses.
6. Memory From Chat History. Let Claude build a working memory of your past conversations so it stops asking you the same context questions. This is the feature that makes Claude feel like it actually knows you.
7. Search and Reference Past Chats. Even without full memory, Claude can search your old conversations and pull the relevant one back in. Stop re-explaining the project you discussed last week.
8. Conversation Export. Pull a full conversation out as a file for your records, your team, or your own archive.
9. Thumbs Feedback. The up and down buttons are not decoration. They are how you tell Anthropic what worked. Use them, and your model experience improves over time.
Layer 2: Projects (The Workspace Most People Skip)
10. Projects. A dedicated, walled-off workspace that isolates files, instructions, and conversation history to one body of work. This is the single most underused feature in Claude.
11. Project Instructions. A persistent system prompt scoped to one Project. Define the role, the standards, and the format once, and every chat in that Project inherits it.
12. Project Knowledge Files. Drop your reference docs, brand guides, and source material into a Project. Claude reads from them in every conversation without you re-uploading.
13. Project-Scoped Memory. Memory and past-chat search respect Project boundaries, so your work project's context never bleeds into your personal one.
14. One Project Per Hat. The real unlock: stop running everything in one endless chat. Spin up a Project for each client, each course, each product. Context stays clean and answers stay sharp.
Layer 3: Artifacts (Claude as a Builder)
15. Artifacts. Ask for something interactive and Claude builds it in a side panel: a tool, a dashboard, a calculator, a web app, a diagram. Runnable, not just described.
16. Persistent Storage in Artifacts. Artifacts can now save and retrieve data across sessions. That means real journals, trackers, and tools that remember your entries between visits.
17. Claude Inside Your Artifact. An artifact can call Claude's API directly, which turns it into an intelligent micro-app with AI running inside it. People call this "Claude in Claude."
18. MCP Connections From Artifacts. Artifacts can reach external services through connectors, pulling live data into the tool Claude just built you.
19. Sharing and Publishing Artifacts. Build a tool once, share the link, let anyone use it. Your one-off becomes a product.
20. Remixing Community Artifacts. Browse what others have built and remix it for your own needs instead of starting from a blank screen.
21. SVG and Diagram Generation. Ask for a flowchart, an architecture diagram, or an illustration and get clean editable vector output, not a blurry image.
Layer 4: Connectors and the Model Context Protocol
22. MCP Connectors. The Model Context Protocol is the open standard that lets Claude connect to outside tools and data through one consistent interface. This is how Claude stops being isolated and starts being plugged in.
23. Per-Conversation Connector Control. Connectors are enabled per conversation, not globally. That means you decide exactly what each chat can reach, which is both safer and cleaner.
24. The Connector Catalog. Hundreds of official and community connectors now exist for the tools you already use. Browse before you build; someone probably made it.
25. Document and File Connectors. Point Claude at your file storage and let it read, search, and reason over your real documents instead of pasted snippets.
26. Communication Connectors. Connect your messaging and email tools so Claude can pull a thread into context. Reading is safe and instant; sending should always stay behind your explicit approval.
27. Calendar and Task Connectors. Let Claude see your schedule and your task list so "what should I focus on today" actually has an informed answer.
Layer 5: Cowork (The Autopilot Layer)
28. Claude Cowork. A desktop tool that grants Claude access to a folder on your computer where it can read, edit, and create files, run multi-step tasks, and coordinate sub-agents. This is Claude working alongside you, not just answering you.
29. Folder Architecture. Cowork works best when you design the folder you give it: clear structure, clear naming, clear instructions. Setup quality determines output quality.
30. Scheduled Tasks. Tell Cowork to do something on a schedule and it runs when nobody is watching. A morning brief, a weekly report, a daily cleanup, all on autopilot.
31. Live Artifacts. Persistent interactive dashboards that live in their own tab, refresh with current data from your connectors each time you open them, and keep version history. A self-updating competitive tracker or morning brief.
32. Sub-Agent Coordination. Cowork can split parallelizable work across multiple sub-agents and stitch the results together, so a big job gets done in parallel instead of in sequence.
33. Project Instructions in Cowork. Just like web Projects, Cowork takes standing instructions so it knows your rules before it touches a single file.
34. Backup Before You Run. The honest setup tip: Cowork can modify files autonomously, so back up the folder before you hand it the keys. Power demands a safety net.
Layer 6: Skills and Plugins (Teach Claude New Tricks)
35. Skills. A skill is a packaged set of instructions and reference material that teaches Claude how to do a specific job to a professional standard. Think of it as a recipe card Claude follows every time.
36. Plugins. A plugin bundles multiple skills, connectors, slash commands, and sub-agents into one installable package. If a skill is a recipe, a plugin is the whole cookbook plus the kitchen.
37. The Plugin Marketplace. There is now a marketplace of ready-made plugins covering sales, legal, marketing, finance, support, and more. Install instead of build.
38. Build Your Own Plugin. Package your own workflows so a new teammate installs one thing and Claude already knows your house style. This is how teams scale their best practices.
39. Slash Commands. Trigger a saved workflow with a single command instead of retyping a long prompt. Your most-repeated tasks become one keystroke.
Layer 7: The Agentic Tools (Claude in Your Apps)
40. Claude in Chrome. A browsing agent that can navigate, read, and act inside your browser. Research, fill, compare, and summarize across live pages.
41. Claude in Excel. A spreadsheet agent that builds models, cleans data, writes formulas, and explains what a messy sheet is actually doing.
42. Claude in PowerPoint. A slides agent that drafts and edits decks, so you start from a real first draft instead of a blank template.
43. Claude Design. A canvas with design tools that Claude drives from your chat, so you iterate on a visual by talking instead of clicking.
44. Claude Code. An agentic coding tool that works from your terminal or desktop and writes, edits, and runs code across your whole project. And it is no longer just for developers; people use it to build entire live products end to end.
45. Claude Code Remote Control. Kick off and steer coding tasks from your phone. Your dev environment becomes something you manage from anywhere.
Layer 8: The Power-User Mechanics
46. The Code Execution Tool. Claude can run actual code to analyze your data, crunch numbers, and produce charts, then hand you the file. This is how you turn a CSV into an answer.
47. File Creation. Claude can produce real Word documents, presentations, spreadsheets, and PDFs you can download, not just text you have to reformat yourself.
48. Vision and Document Analysis. Upload screenshots, diagrams, photos, and PDFs and Claude reads them. High-resolution vision means it can actually parse a dense chart or a full page.
49. Deep Research Mode. For big questions, this mode runs many searches across many sources and returns a synthesized report instead of a single shallow answer. Use it when one search isn't enough.
50. Extended Reasoning. For genuinely hard problems, let Claude think longer before answering. The harder the question, the more this matters, and most people never give it the chance.
How to Actually Roll This Out
Fifty features is overwhelming if you try to adopt them all on a Monday. Don't.
Here is the order that works. Spend day one on Layer 1, the foundation settings. Preferences, a custom style, memory on. That alone changes every conversation you have from then on.
Day two, build one Project for your most important work and load it with instructions and reference files. Live in it for a few days.
Day three, build one artifact you will actually use and connect one data source through MCP. Feel what it's like when Claude can reach your real information.
After that, add Cowork or a plugin only when you hit the ceiling of what the simpler layers can do. The mistake is reaching for the most advanced feature first. The order is foundation, then workspace, then builder, then autopilot.
The Honest Truth About Setting Up Claude
A great setup will not save a vague request.
All of these features amplify clarity. If you know exactly what you want, the right Project plus the right connectors plus the right instructions will get it for you faster than you thought possible. If you don't know what you want, fifty features just give you fifty ways to get a vague answer.
So the real work is the same as it always was: get clear on what you're trying to do. The setup is what makes the clarity pay off.
Most people will read this list, nod, and keep typing one question into an empty chat box.
The small group that actually sets up even half of these will quietly start producing work that looks like it came from a whole team.
That group is still small. The door is still open.
The only question is which group you'll be in by the end of the week.
If you found this useful, follow me @eng_khairallah1 for more AI content like this. I post breakdowns, courses, and tools every week.
hope this was useful for you, Khairallah ❤️
