Claude Code has now become super cheap because they've brought it onto the $20 pro plan and the $200 max plan instead of API usage, and that's why it's become my go-to coding agent. You've probably watched tons of tutorials on it, but here are some amazing features and tips that you might have missed. And I promise you that knowing about them is going to make you 10 times more productive with Claude Code. And at the end of the video, I'll also let you know the key method that you can use to push Claude Code to its maximum power. Okay, so the very first feature that I want to show you and that I don't think you've seen before is making custom workflows or you might even say custom agents in Claude code. Let me just fire up Claude so you can see what I'm talking about. So if you head into your Claude and type these slash commands, you get multiple commands that allow you to take advantage of Claude's amazing capabilities. But you'll see here that I have this command called search web. And if you go into Claude code, you won't have this here. That's because this is a custom/comand I've made myself. If I go ahead and ask it to do something like for example, how do I use the custom claw feature in claude code, you'll see that it's going to search the web for that term and get me the most to the point solution to my problem. And you can see that it has returned detailed output on how to use this custom command feature. So what are these custom commands? These are essentially reusable prompts with arguments if you choose to add them that allow you to build custom workflows Claude Code can execute for you. That way you don't need to repeat the same prompt again and again. You just create these custom/comands once and Claude Code will do it for you. Now, how do you create these custom/comands? Let me show you. First, you're going to go to your base directory and paste this command which will create a commands folder inside Claude Code. That folder is where you'll create your custom workflows. If I go into that folder and open it up, you can use any text editor you like. I'll just open it in cursor. You'll see we have this search webb.md file. These MD files are the slash commands that will show up in cloud code. For example, you can see I've laid out this prompt here telling it to search the web for the given argument. This is how you define arguments, the dynamic part of the prompt that changes every time you run it. And I've asked it to return the most relevant information about the search query. Now, if we want to create another custom workflow, for example, a codebase explainer, we would create another file and name it explain codebase.md. In here, we'll write out our custom prompt. For instance, we're going to ask it to scan the whole codebase and explain it to us in an easy way that helps us understand how it functions and how everything is connected. This becomes a reusable prompt and every time I call this custom command, it will just run that prompt automatically. So if I go back since I've updated the commands folder, I need to restart cla code first so it can recognize the new command. And now if I type the /ash command, you'll see that we have our explain codebase tool which will explain the whole codebase and how it functions. Before we continue to the second feature, let me show you another cool thing. You've probably heard that you can now use clawed code inside cursor or windsurf. And if you're using this feature, there is one thing you definitely need to do. Otherwise, you're going to have a frustrating experience. First, for those who don't know, let me show you how to get claude code inside your terminal. You open up your terminal inside cursor and just type the claud command. This initializes claude code and automatically recognizes that you're inside cursor. Then installs the claude code extension. Now you can see it's been installed. What we need to do is press command plus escape which will open up claude code for us inside cursor. You can see that claude code is now initialized in cursor and it's going to receive everything. All the errors from cursor can also be passed to claude code. Now, here's something you probably didn't know. If you're using cloud code and want to write multi-line prompts or structure your prompts properly, you're probably facing a major issue. For example, if I want to make a website and give it specifications on the next line, when I press either shift plus enter or just enter, you'll see that it starts running the prompt immediately. This is not the usual behavior you'd expect inside cursor. So, how do you get that multi-line editing feature in claude code? There's a special command that's only available in cursor. It's not available in your regular terminal. Once you enter this command, it sets up cursor terminal key bindings, allowing you to write multi-line prompts properly. So, now when I type, I can move to the next line and give it different details without accidentally triggering the prompt. Oh, and if you're enjoying the content we're making, I'd really appreciate it if you hit that subscribe button. We're also testing out channel memberships, launch the first tier as a test, and 50 people jumped in immediately. The support's been incredible, so we're thinking about launching additional tiers. Right now, members get priority replies to your comments. Perfect if you need feedback or have questions. Now, the second feature that I want to tell you about Claude Code is that you can actually use MCP servers with it. And I don't know if you've been using different MCP servers in cloud code, but it's a gamecher. If you haven't, I'll show you a great example of why it's useful and how you can set them up yourself. So, if I use the /ash command and open up the manage MCP server section, you'll see that the Sentry MCP has been connected to clawed code. You can clearly see in this video, Sentry is a platform where you manage issues in your code. It's an amazing platform that software developers have been using for a long time and its MCP server takes it to the next level. We also made a video on it. You can see in the video that it can fetch issue details about a particular project. Using those details, your AI agents, whether it's clawed or cursor, gain much more insight for solving that issue and debugging it. It even has its own AI powered agent that helps you solve these problems. This is why using MCP servers in claude code is so important. It adds so much functionality and Sentry isn't the only one. You can add so many more amazing MCP servers. I'll leave some links down below to our videos on other MCP servers that you should definitely check out. Now, the next question is how do you install these MCP servers for Claude Code? I'm going to give you the most straightforward method so that the existing JSON configuration you're using with Cursor and Windsurf also works with Claude Code. First, you'll want to exit claude code so you can add your MCP servers. Then you're going to paste in a command like this. So what's happening in this command? You write claud followed by MCP. And then since we're adding JSON in this method, we use the add JSON method. Next comes the name of the MCP server, which in our case we want to set as sentry. This is the name you define. It's not hard-coded, so you can change it, but as a best practice, it's usually a good idea to use the same name you see in the configuration. For example, you'll see that this is the JSON for the Sentry MCP. You just copy it specifically this part of the JSON and go back and put that after the name inside single quotes. So you can see first we wrote the name we wanted which was sentry. Then we added a single quote, pasted the highlighted part of the JSON from earlier and closed it off. If you run this command, it will add the Sentry MCP to Claude code. To verify this, just open claude code again, use the MCP command, and you'll see that the Sentry MCP is available and ready for you to use. Another useful thing about slash commands is that you can use MCPS directly with them. For example, if I type the slash command with MCP, you'll see that the Sentry MCP provides us with two commands that include arguments we can use. One is find errors in a file, and the other is fix issues with SEIR, which is their AI agent I mentioned earlier. When I enter this command, I need to provide an issue ID as an argument. This tells it exactly what issue to fix with Seir. This essentially turns it into a programmable tool that you can call whenever you need it. Now, before I tell you how to push Claude Code to its maximum limit, let me tell you about the final feature. Claude Code now has a new option called plan mode, which allows you to plan out your whole project or idea. It's basically a way to brainstorm anything with clawed code and you can access it by pressing shift plus tab to switch modes. As you can see, plan mode has been turned on. You can also switch into auto accept edits mode. This mode will edit files without asking for your permission. It's definitely better when you want to give full autonomy to Claude Code and just let it do its thing. When you're in plan mode, you can type out your ideas, whatever you want to brainstorm, and it will flesh out your thoughts, give you more clarity, and even allow you to put it into the claude.mmd file, which you can initialize using a slash command. This file is essentially a rule file for Claude code, and Claude will follow it. You can include anything in it, codebased, documentation, specific rules, and it will help guide your workflow. There's also install GitHub app, which is an amazing GitHub app. We've featured it in another one of our videos where we discussed the Claude SDK which now allows you to embed Claude code into AI applications. Yeah, you heard that right. You can actually put Claude code into apps and program it. So definitely check that video out as well. Now, here's how you push Claude code to its maximum limits. Well, you just need to use one keyword and it's called Ultraink. If you think I'm just making this up, you can actually see that Claude themselves have referenced this, that using the keyword ultrathink will push Claude code and the models it uses to maximum power. You should only use it for your hardest problems that Claude can't solve in its normal state. That brings us to the end of this video. If you like the Sentry MCP, do check out our video on that as well. If you found these methods helpful and want to support the channel, you can use the super thanks button below. It really helps us keep making tutorials like this. Thanks for watching and I'll see you in the next