The Claude Glossary: 50 Words Every Power User Knows

I remember the first week I really used Claude. People kept throwing around words like “MCP,” “artifact,” and “context window,” and I nodded along while quietly opening a new tab to Google every single one. If that sounds familiar, you’re going to love this one.

I just came across a post from an AI professional who says they spend about four hours a day inside Claude. The creator put together a vocabulary checklist with a bold claim: the quickest way to tell if someone has actually mastered Claude is whether they know all of these words. They also make it clear they aren’t paid by Anthropic. They just use the tool a lot and think you should too.

I went through the whole list and found it really useful as a self-check. Below, I’ve organized the author’s glossary into five numbered sections. Each term gets a short explanation plus a few notes from me on why it matters in practice. Keep a mental tally of how many you already know.

1. The Basics: Your Starter Vocabulary

These are the foundation words. If you skip them, the rest of the list won’t make much sense.

  • Add-on: Claude working inside Google Workspace, so you don’t have to leave your docs or inbox.
  • Agent: AI that takes actions for you instead of just answering questions.
  • Anthropic: the company that builds Claude.
  • Artifact: the things Claude builds for you, like motion graphics or slides, shown in a separate panel.
  • Ask me first: the author’s swap for the word “never” in prompts. Instead of banning an action, you tell Claude to check with you before doing it.
  • AskUserQuestion: a clickable question form Claude uses to gather what it needs from you.
  • Auto: giving Claude permission to act without asking you each time.
  • Brief: a summary you carry into a new chat so you can switch without losing context.
  • Browser use: Claude clicking around the web for you.
  • Cache: Claude’s saved copy of your chat.
  • Chat: the one box where everything starts, whether you’re writing code, designing, or building.

My favorite tip here is “Ask me first.” Telling a model “never do X” can backfire because it either gets too cautious or ignores the rule. Asking it to check in first keeps you in control without slowing things down.

2. Claude Features: What It Can Actually Do

Once the basics click, this section shows how far the product has grown beyond a simple chat window.

  • Claude Code: build websites and apps. The author recommends using the desktop app.
  • Claude Data: your company database, available inside Claude.
  • Claude Meetings: Claude joining Google Meet to assist you.
  • Claude Motion: animated videos made inside Artifacts.
  • Claude Sheets: spreadsheets, like Google Sheets or Excel.
  • Connector: outside apps you can use inside Claude.
  • Context bloat: when there’s too much information for Claude to keep track of.
  • Context window: the maximum amount of information Claude can handle at once.
  • Cutoff: the date of the model’s most recent training data.
  • Dashboards: data turned into live, moving charts.
  • Done means: a preview of what the finished task should look like.

The pair to remember is context window and context bloat. When a long chat starts to feel off, it’s usually bloat. That’s when a “Brief” from section one helps: summarize, open a new chat, and carry on.

3. Tools: The Knobs and Buttons

This is where everyday habits make the biggest difference.

  • Edit: the pen icon on a prompt you’ve already sent. Fix the prompt instead of stacking corrections.
  • Effort: how hard Claude thinks before it answers.
  • Em-dash: according to the author, the biggest sign that text was written by AI.
  • Enterprise: the subscription for big companies.
  • Fable 5.2: the latest and smartest model, per the post.
  • GitHub: like Google Drive, but for code.
  • Goal: what you want. The author says it should be in every prompt.
  • Haiku: Claude’s smallest model. The author’s advice is blunt: don’t use it.
  • HTML: the author suggests asking for HTML files instead of images, much like a Word file.
  • Instructions: found in Settings. The author’s take is to delete them.
  • Max: the $100-a-month subscription plan.

A couple of these are clearly the creator’s personal opinions, like skipping Haiku and clearing your custom instructions. I’d test them on your own work before you commit. The Edit tip, though, is gold. Rewriting the original prompt usually gets a cleaner result than five follow-up messages.

4. Plans and Prompting: Working Smarter

Next up is how to structure your work and pick the right setup.

  • MCP: a way to plug your own apps into Claude.
  • Memory: context about you that carries across chats.
  • One Claude: the change where the separate Cowork tab went away.
  • One-shot: one prompt that gets you one finished result.
  • Opus 5.5: the author’s pick for daily tasks, set to High effort plus Auto.
  • Plugin: a bundle of Claude skills installed together.
  • Pro: the basic $20-a-month plan.
  • Project: shared context across all the chats inside one project.

If you only try one thing from this section, set up a Project. Putting your brand guidelines, past work, and recurring instructions in one place saves you from repeating yourself in every chat.

5. Mastery: The Power-User Layer

Last comes the stuff that separates casual users from people who really live in the tool.

  • Team: the team subscription, $20 a seat.
  • Skill: a pre-written instruction you can run like a command.
  • / command: how you activate skills, for example “/goal”.
  • System prompt: the underlying instructions Claude follows.
  • Vibecode: building on vibes and pretty much forgetting the code exists.
  • Whiteboards: like Miro, Claude draws out your project visually.
  • Thumbs: the thumbs up and down under every answer, used to give feedback.

Skills and slash commands are where the real time savings live. Write a prompt once, save it as a skill, and run it in seconds every time after that.

How to Use This List

Here’s how I’d put this glossary to work:

  1. Count how many terms you could explain to a friend without looking them up.
  2. Pick three you don’t know and try each one this week.
  3. Start every prompt with a clear “Goal” and a “Done means” description.
  4. Use Edit instead of piling on corrections.
  5. Turn your most repeated prompt into a Skill.

A shared vocabulary does more than make you sound smart. Once you know what a feature is called, you start noticing when to use it, and that’s when the time savings show up.

The author’s own closing note says they spend four hours a day on Claude and think you should too. Whether you go that far or not, this list is a great map of what’s possible. Check out the full LinkedIn post to see the original glossary and test yourself.

Scroll to Top