Key insights

• CLAUDE.md is persistent memory for a tool that forgets everything between sessions. Without it, you repeat yourself every time you start a new conversation.
• Plugins like Context7 solve a real frustration: Claude Code sometimes suggests outdated code because its training data has a cutoff date. Context7 fetches current documentation on demand.
• The /insights command generates a full usage report about how you actually work with Claude Code. Most users never discover it exists, but it can reveal habits worth changing.

In Brief

Tim Ruscica, software engineer and YouTuber behind the Tech With Tim channel, walks through five practical changes that can make Claude Code dramatically more useful. The tips cover where to run Claude Code, how to give it a permanent memory, which plugins to install, how to connect it to GitHub, and which commands most beginners never discover. Tim calls the combined effect a 10x productivity improvement, and the changes take only a few minutes to set up.

This article explains what each tip means and why it works. For visual, step-by-step demonstrations of every setup, watch Tim's video directly. The article explains the concepts; the video shows them in action.

---

Tip 1: Use Claude Code inside a code editor

Claude Code is a tool made by Anthropic that lets you build software by having a conversation with an AI. Instead of writing code yourself, you describe what you want and Claude Code writes it for you.

You can run Claude Code in two different places. The first is a terminal, which is a text-based window on your computer where you type commands to control things. Instead of clicking buttons, you write instructions. The second is inside a code editor, which is a program designed for writing and reading code, with helpful features like file browsers, syntax highlighting (color-coding that makes code easier to read), and extensions.

Tim recommends using Claude Code inside a code editor like Cursor, VS Code, or Windsurf. The reason is simple: when Claude Code creates or changes files, you can immediately see what it is doing in the file browser next to your conversation. Without that, you are working a bit blind: you get the output but cannot easily see the full picture of what was added, changed, or deleted. Installing Claude Code as an extension inside your editor takes about one minute and gives you much better visibility over your project.

---

Tip 2: Create a CLAUDE.md file

This is the single most important habit for any Claude Code user. Here is the problem it solves.

Claude Code has no memory between sessions. A session is one conversation with Claude Code. When you close it and open a new one, it starts completely fresh. It can read the files in your project, but it has no idea what rules you set last time, what decisions were made, or what conventions your project uses. You would have to explain everything from scratch, every single time.

A CLAUDE.md file solves this. It is a plain text file you create and place in the root folder of your project (the main folder that contains everything else). The name must be written exactly as CLAUDE.md, in all capitals. Every time you start a new Claude Code session, it automatically reads this file before doing anything. Think of it as a permanent sticky note on your desk that your assistant reads at the start of every shift.

What should you put in it? Tim recommends including:

• The tech stack your project uses (the programming languages and tools it is built with), so Claude Code does not randomly suggest a different language
• Development commands (the specific instructions needed to run, build, or test your project)
• Code conventions (how you prefer code to be written: naming styles, formatting rules, and so on)
• Project structure notes
• Any API references (an API is a way for programs to talk to each other, and references are the documentation that explains how to use them)

Tim also suggests telling Claude Code to update this file as the project grows. That way, important decisions get captured automatically, and the file becomes more useful over time.

---

Tip 3: Install plugins

A plugin (also called an extension or add-on) is extra software that adds new abilities to a program. Think of it like installing an app on your phone: your phone already works without it, but the app gives it new skills. Claude Code supports plugins that extend what it can do.

Tim recommends three plugins every beginner should install.

Playwright

Playwright is a tool for controlling a web browser automatically. Normally, when you build a website, you test it by opening it in a browser and clicking around yourself. Playwright lets Claude Code do that work for you instead. It can open your website, navigate to specific pages, click buttons, fill in forms, and check whether everything works, all without you touching the mouse. This is called browser automation, and it saves a significant amount of testing time.

Superpowers

The Superpowers plugin adds structured "skills" to Claude Code: pre-programmed ways of approaching common tasks. These include brainstorming skills (which Claude Code runs automatically before creative work like designing a feature), parallel agent dispatching (running multiple AI tasks at the same time), and systematic debugging procedures (finding and fixing errors in your code). When Tim asked Claude Code to brainstorm a landing page concept with this plugin installed, Claude Code immediately activated the brainstorming skill on its own and produced a structured plan.

Context7

This plugin solves one of the most frustrating problems with AI coding tools. Claude Code was trained on code and documentation that existed up to a certain point in time. Programming libraries and frameworks (the ready-made building blocks developers use) update frequently. Code that worked two years ago might be deprecated (outdated and no longer recommended) today.

Without Context7, Claude Code sometimes suggests deprecated modules or old approaches that no longer work in modern versions of a library. Context7, built by Upstash, fixes this by fetching the most current documentation at the time you ask a question. Claude Code can then give you answers based on today's version of whatever tool you are using, not a version from a year and a half ago.

You can install plugins globally (available in all your Claude Code sessions on your computer) or at the project level (only available when you are working inside one specific project). Tim recommends installing these three globally, as they are useful in every project.

---

Tip 4: Connect the GitHub MCP server

This step connects Claude Code to GitHub, which requires a brief explanation of two things first.

What is Git? Git is a system that saves snapshots of your code as you work. If something breaks or you make a mistake, you can go back to an earlier snapshot. Think of it as an unlimited undo button with a full history of every change you ever made.

What is GitHub? GitHub is a website where you store your Git snapshots in the cloud. It lets you access your code from any device, share it with teammates, and use it to deploy your project (publish it so others can access it on the internet). Every project on GitHub is stored in a repository (often called a repo), which is a folder containing all your files and their complete history.

Normally, connecting your local code to GitHub requires several manual steps in the terminal. But with the GitHub MCP server installed, Claude Code can do all of that for you: create a new repository, push your code to it, and more.

What is MCP? MCP stands for Model Context Protocol (a standard set of rules, created by Anthropic, that lets AI tools connect to external services). An MCP server is a piece of software that follows those rules, acting as a bridge between Claude Code and another tool. The GitHub MCP server is that bridge to GitHub.

To set it up, Tim walks through these steps in the video:

1. Go to the GitHub MCP server setup guide
2. Create a Personal Access Token on GitHub (a special password that allows Claude Code to act on your behalf). You only need to grant it access to repositories.
3. Run the setup command in your terminal with your token, adding a --scope user flag so the server is available across all your Claude Code sessions, not just one project.
4. Verify it worked by typing /mcp inside Claude Code. You should see GitHub listed as a connected server.

Once connected, you can simply ask Claude Code in plain English: "Create a new repository on GitHub and push my current code to it." Claude Code handles everything.

Tim also mentions DBHub, another useful MCP server that gives Claude Code direct read and write access to a PostgreSQL database (a common type of database used in web applications). With DBHub, Claude Code can look at your actual data, run queries (questions you ask a database), and write better code because it understands the real structure of your data.

---

Tip 5: Three commands most beginners miss

/model

Typing /model inside a Claude Code session opens a menu where you can switch which AI model you are using. Claude Code supports several models, each with different strengths:

• Sonnet: the best balance of speed and capability, recommended for most tasks
• Opus: more powerful for complex problems, but slower and uses more of your monthly quota
• Haiku: fast and lightweight, good for simple or repetitive tasks

/compact

Every AI conversation has a context window, which is the amount of text the AI can hold in memory at once. As a conversation grows longer, older parts start to get pushed out, and the AI begins to forget earlier context. The /compact command tells Claude Code to summarize the entire conversation into a compact version, freeing up space so you can keep working without starting a new session. Claude Code also does this automatically when it detects the context window is getting full, but you can trigger it manually at any time.

/insights

This command generates a full web page report about how you use Claude Code. The report shows how many sessions you have started, how many messages you have sent, what kinds of tasks you tend to work on, how Claude Code is performing for you, and suggestions for improving your workflow. Tim describes it as a meta-analysis of your own tool usage, and says reading through it revealed habits he did not know he had. Most users never discover this command exists.

---

Why these five changes matter

Each of these five tips is what Tim calls "low lift": they take little time to set up but deliver significant results. None of them require deep technical knowledge.

Together, they transform Claude Code from a bare-bones AI chat interface into a system with persistent memory, specialized skills, real-time documentation, external service connections, and self-reporting. The difference between Claude Code without these changes and Claude Code with them is substantial, even for a complete beginner.

---

Glossary

---

Sources and resources