Getting Started with Claude Code: The Ultimate Beginner’s Guide (Not Just for Coding)

Key Takeaways

Clawed code is more powerful than most people realize. It's not just for developers. It can batch process files, build automations, and solve problems that regular chat GPT or Claude simply can't handle. [Music] But here's the thing. Most people think the command line is too intimidating to try. [Music] This is your ultimate getting started guide for Claude Code for anyone that wants to unlock its power. No coding skills required. Let's go. Hey, Alex here from Grit AI Studio. If you came for the quick start, tap the timestamps and jump right in. But I'd urge you to watch the whole thing. I'll show you why this isn't only for engineers and how it unlocks a lot more than just write some code. Even though their website also targets the typical use cases and roles relating to development, as you'll learn, there's so much more potential than just coding. So, Claude Code's been around for a while and helped popularize the idea of a terminal native agent, an assistant that can actually edit files, run commands, and make commits with your approval with a huge benefit of being able to run available tools in the command line. I'll come back to the power of the command line, but that part has maybe also been what's prevented a lot of people from touching claude code because let's be honest to most non- tech people the command line or terminal can be quite intimidating. Fortunately, today it also shows up in a friendly VS Code extension and more recently on the web if you don't want the terminal. I've used it for coding since May. But the real eye openener for me came this summer when Anthropic showed how they use it for much more than coding. Docs, release notes, cleanups, packaging, real actions on your projects, not just chat. That's when I went all in on non-coding use cases. As the anthropic team shared, agentic coding tools like claude code helps developers accelerate workflows, automate repetitive tasks, and tackle complex programming projects. But they also discovered a range of new applications beyond coding. Marketing teams using it to generate hundreds of ad variations and data scientists creating complex visualizations without knowing JavaScript. Essentially, they realize that this is dissolving the boundary between technical and non-technical work. Turning anyone who can describe a problem into someone who can build a solution. At Grit AI, we're obsessed with helping more people solve their own problems with AI. Our clients keep asking, "How do we do X or Y that chatbots struggle with? Batch processing, folderwide analysis, or tasks that simply blow past a chat window's limits." We realized that claude code was the missing piece. Aentic coding tools help teams build custom automations that would traditionally require dedicated developer resources or expensive software. A few weeks ago, we got Claude skills. And now a lot of the influential voices online have also realized the power of Claude code for use cases beyond coding. So I decided it was time to put together a proper nononsense guide to claude code. Let's jump in. So what exactly is Claude Code? Claude Code is an assistant that can do work with your files. It's called a CLI based coding agent. CLI just means command line interface. A simple window where you type a command and press enter. The terminal is that window. With Claude code, you say what you want in normal language. It proposes the exact changes, asks for your okay, and then does the boring parts for you. It's scoped to the folder you're working in. It clearly started with a focus on devs, and the value prop for devs is not that difficult to grasp for tech people. CLI based coding agents are powerful because they have unrestricted access to the developers entire toolkit. They can directly execute git commands, run build systems, invoke package managers, and manipulate files without sandbox limitations. They integrate seamlessly into existing developer workflows since developers already live in the terminal and can compose operations by chaining commands together using the full Unix ecosystem. Now, that's a little bit technical, but let's try to explain this in slightly more plain English. Coding agents like Claude Code use bash tools or command line tools combined with their own specialized tools to do things like search through files, read file content, exploring directories, searching and processing text. Think cat, glob, gp, and so on. for those that are a bit technical. All of this, of course, on top of being powered by really smart language models that excel in writing code. They can also start and stop processes in the system, run package managers to install packages, or build and run code that is generated. If they have network access, they can run network operations like fetching remote resources. Think curl or VGA for those that are a bit more technical. Now, as anthropic discovered, there are so many use cases also for non-development tasks. For non-development tasks, CLI based agents are incredibly powerful because they turn your computer into a programmable workspace for any file-based task. They can orchestrate entire workflows like extract data from these 50 PDFs, analyze it with Python, generate charts, and compile a Word document. All automatically using the tools already on your system. Unlike web-based AI that requires uploading files and has limited capabilities, CLI agents work directly on your local files with privacy intact and can handle tasks like batch processing thousands of images, web scraping with curl, converting file formats, organizing documents, generating reports from spreadsheets, or creating entire presentations. They're essentially a universal automation layer that can use any command line tool to solve real world problems. data analysis, content creation, media processing, file management, not just coding. So instead of claude code, they should have named it Claude agent. Since Claude code was launched, others have launched similar agent in their terminal tools. Claude code was the OG, but now you can do most if not all of what we are talking about here today in any of these tools. Codeex CLI is the coding agent by OpenAI. Gemini CLI is the equivalent offering from Google. GitHub copilot CLI is another option and even cursor have launched their own CLI agent. The advanced features started in Claude but now most of these offer stuff like slash commands, sub aents and hooks, all of which we will cover in future videos. Okay, enough talk. Let's get started. To get up and running, you need a paid plan. You can start with the pro plan at 20 bucks per month, but if you end up using this a fair bit, you'll want to move over to one of the max plans. It might seem pricey at $100, but it's honestly well worth it once you start using it every day. Either way, start on the pro plan to get a feel and take it from there. With your account set up, all you have to do is open up your terminal or command line. If you're a bit technical, you might already have NodeJS installed on your computer. If not, you need to first download and install it. Head over to the Node.js page. Select your platform. Mac in my case. You have several options. If you're unsure, just go with the default choices. Just follow the steps in the guide here. With node installed, we can make use of the node package manager, npm, to install cloud code with this simple command here. Just copy and paste it into your terminal window and hit enter. Make sure you use the official package here. And also be aware that there are fake or malicious packages out there. So when you or Claude installs npm packages, just make sure they are the official ones. Now, all you have to do is enter a folder where we want to run our project. It's highly recommended to run cloud code inside a project folder. Eventually, you want to use source control such as GitHub, but don't worry about that right now. To get started, just create a folder and move inside it with the change directory command in the terminal cd and the path to the folder. Here, I move into a folder called dev and a subfolder called my project. Now, I know this can be a bit intimidating, but just follow the steps and you'll be up and running in no time. Then, we can type Claude and hit enter to fire up Claude code. You'll be asked a few questions and prompted to log in on first use. Once logged in, your credentials are stored and you won't need to log in again. You'll be asked if you trust the folder you're in and once you confirm, you're ready to roll. Now you can start chatting with Claude code the same way you talk with Claude or ChatgPT. Claude Code can read the files in your directory if you have any and it also has access to its own documentation and can answer questions about its features and capabilities. Now there are some alternative ways to install cloud code if you struggle with the npm approach. Although these native installers are in beta, you can definitely try it out. You can use Brew on Mac or on Windows for instance. You can either use this PowerShell command or this command line command. Now we're in an empty project. So this command won't really give us much value, but let's try it anyways. Claude code reads any files to explore the codebase. But in our case, we have an empty project. So there's nothing much to explore really. It will ask for permission when running certain tools and eventually it'll report back that this is an empty project. We can essentially ask anything, but of course the fun part is to ask it to build stuff. So let's start with a simple Python script that just prints my name when we run it. If you haven't installed Python, that's also smart to install. Typically, unless we tell you something different, Claude will solve problems with code in popular languages like Python or maybe TypeScript. If you're missing libraries to run this, you can also ask Claude to help you install it. This is just a super simple example. Of course, you can ask any question and Claude will answer or build a solution if that's what's required to answer. Tell Claude what you want to build in plain English. It'll make a plan, write the code, and ensure that it works. We'll come back to all the advanced features and ways to really get the most out of Cloud Code in future videos, but for now, let's look at a few simple examples to get you started. So, we're keeping things simple today, and I'll cover tools and workflows in a future video. But one important tip for you is to regularly use the slashclear command to clear the context window. When you switch tasks, move from one feature to another, and so on, make a habit of clearing the context. First, let's ask Claude to actually code something for us. Let's drop in a CSV file with some data in the folder here. Then, ask Claude to analyze the data, build a nice interactive dashboard for us. We run this in something called accept edits mode by clicking shift and tab. This means we're telling Claude it's okay to make changes to the files in our folder which we need if we're writing code and don't want to review every single change that it's making for us here. Claude will ask us when it tries to run tools. I'll accept these so that it can actually run and test things. Unless we're specific, Claude will make the choices for technology and so on for us. Like in this case, it's using Python, Flask, and something called Plotly for the charts. Claude gives us a summary and instructions for how to run this. Or we can simply ask it to run it for us. It caught some errors that it fixed. And there we have it. Now we can give feedback, add functionality, and expand as we like. A bunch of cool possibilities here. A nice trick is to take a screenshot and paste a picture in. We'll use CtrlV. And now Claude can use our picture as context. Or maybe we want dark mode instead. There we go. It's that easy. Of course, for proper coding, we should follow some best practices. Make a plan first, use version control on our projects like Git, and set up rules and so on. We'll cover that in another video. Today, we're keeping things simple for everyone to follow along. Let's look at some cases beyond coding. Let's say you have a folder with PDFs that you want to extract something from. We simply explain to Claude what we want and it'll get to work. Create a summary of the reports in the reports folder. You can reference files using the at symbol. Provide it in a table format by date and key themes for comparison. We'll tell it to use something called sub agents as each agent has their own context window. PDF sizes matters, but if you hit the wall on that, simply ask Claude to split up the files, do the analysis in steps, and so on. We can now ask for the output in an Excel file. What's really cool is that Claude will build the tools that it needs to solve the challenge. Once built, it can be reused in that project, which is really part of the magic here. You can use this for a bunch of use cases. Maybe you have a folder with receipts and you want to automate the process of filling in expense reports. Just spell it out to Claude. The thing is, we can use Claude code for tasks that you're not able to solve in regular Claude or ChatgBT. If it involves multiple files, if it involves repetitive tasks, if it involves multiple steps and so on, Claude code can very likely help you out. So, here's my power workflow for literally any task outside coding. Create a folder for your project. Make a file instructions MD and write your step-by-step instructions in plain English. Think how would you tell it to your intern what to do step by step in plain language. Save the file and now ask Claude to follow the instructions. The magic here is that Claude will review, think, and look through its tools. If it's missing something to get the job done, it will try to make the tools itself. You ask for it to download PDF files from a website, it might create a script to fetch PDF files. You ask it to access an API to get some data. It will make a script to access that API. It might then ask you to provide an API key, for instance, to proceed. There are endless possibilities with this approach. [Music] And as a bonus tip, you can of course also ask Claude to help you write out those instructions for even better results. [Music] So again to recap, here's the real magic. You just give clear instructions. Claude code decides what tools it needs. If it must fetch data, it can write a tiny helper script. If you say analyze every PDF in this folder, it will walk the folder, process each file, and produce outputs that you can open. For bigger jobs, it can even delegate to sub agents, little specialists that take on parts of the task, each with their own context window. There's so much power in this, and I'll show a lot more examples in upcoming videos. We'll dive into sub agents, slashcomands, agent skills, hooks, plugins, and MCP connectors to tools like GitHub, Notion, and so on in separate videos. But now you're at least up and running, ready to start solving problems with the help of Claude Code. Before we wrap up, let me just say a few words about safety. The great thing is that you're always in control. Claude asks for permissions before changing files. Just make sure that you keep secrets out of the prompts. Work inside a project folder, not your whole home directory. If using it for coding and you use the web version, your session runs in an isolated environment that you can steer. All right, that's it for today. In the next video in this series, we'll cover the essential commands and workflows for Claude Code and show more cases that we can solve with the help of Claude Code. I hope you found this video helpful. Hit that like button and subscribe for more Claude Code videos. I'll see you in the next one.