I Tested Every Claude Code Feature, These 12 Are the Best

So, I've spent well over 500 hours inside of Claude's ecosystem, chat, co-work, Claude Code. So, what I wanted to do today is break down all of these features, rank them from D all the way up to S tier, and walking you guys through my top 12 in order, and telling you why. So, let's not waste any time and just get straight into the video. All right, so here we go. Every cloud feature ranked, not literally every single one, but there's a lot of them in here. So, real quick, I wanted to preface how I actually ranked these. These aren't ranked by globally which features get used the most or how cool I think they are. These are ranked by how much it changes my actual day-to-day when I'm doing my work. And keep in mind, I am doing a lot of knowledge work and automation. I'm not as much building software or SAS products. So, if you are, you might disagree with some of these. And look how sassy my cloud code was cuz my cloud code obviously helped me build these slides. You will disagree with some. Good. The point being, you value features differently based on the way that you use cloud code, right? So, take my ranking with a grain of salt, but hopefully I can surface some of these features that you may not have known about or may not be using correctly. So, let's just jump right in. Okay, so starting with our D tier, I'm not going to read every single one. You guys can screenshot this or pause this if you want, but I wanted to call out a few that I think are kind of the standouts of this tier. So, these features are all very real and very useful, but it's just like nobody's favorite feature, right? Like clawmd. Obviously, we need that. It's super important to set up and if you don't have that in your project, it's not going to feel very good, but it's just not a very very like cool feature. So, like I said, very useful, but I put this in the D tier. It's kind of just like the groundwork. It's like the bare minimum context life cycle. So, doing things like /clear/compact, resuming a conversation, all useful. Web search, you know, cloud code obviously has its native web search and then you can do your web fetches, which is cool because it will spin off research for you right away. So, once again, just like a very basic feature, but it is still useful. And then we've got some other things in here like file uploads, fast mode, permissions, custom themes is something that you can do. And I actually didn't even know about this until this video. You can see here I just tried out themes. You come in here, you do/ theme, and you can change like auto, dark mode, you can do colorblind friendly, and you can even set up your own custom themes. I threw the IDE extension in here because in cloud code in VS Code, where I like to use it, I typically like to work in the terminal, but you can also use the extension, which is where I started off when I was first learning cloud code. So anyways, that is my tier D. Let's move on to tier C. These are other functionalities or features of Claude that once again I think are great. I just don't use them a ton. So starting off here with voice mode, you can come into Claude and you can do a /voice and this lets you basically just talk. And now as you see Claude is hearing what I'm saying and it is basically dictating for me which is obviously awesome. Or if you want to do it the much cooler way, you just use a tool like Glido and now you're doing it even faster and better. As you can see, boom. Then we also have Co-work. And don't get me wrong, I think co-work is an amazing product, but for me, I just went straight to claude code. And I feel like pretty much anything you can do in co-work, you can do in code. So why not work in a tool that has more capability and just learn that tool really well. Just because it's called Claude Code doesn't mean you have to be writing software and like heavy engineering, heavy coding. You guys have seen me talk about this kind of stuff a lot. The way I use Claude Code as my executive assistant and I've kind of turned it into my AI operating system. So that is why co-work is a Ctier for me. But hey, if you are completely non-technical and you're overwhelmed getting into this whole claude thing and if you want to start with co-work, that's absolutely fine. I also threw in here interactive connectors. When you're in claude and you're doing things like, you know, in your chat or co-work or wherever you are and you want to just throw in connectors, it's really easy to just be able to log into something and automatically have all of the tool calls available rather than having to maybe try to play with setting them up in the terminal. So that's why this is a Ctier. It's still very useful, but I just don't use that as much. The other one I wanted to highlight here was local file access. Obviously, that is a huge benefit of these kind of local terminal coding agents. All right, so now let's move on to our B tier, which are the ones that I think are super super useful, but didn't make my top 30. So, dynamic workflows. So far, I've actually been loving this feature. It definitely uses a lot of tokens, but it processes things in parallel really, really well. And I just made a full video breaking this down if you want to see. I'll tag that right up here. But the reason I gave it a B tier is also because at the time of making this video, it's very new. So, I've been experimenting and I like it, but I haven't yet found out how exactly it works into my day-to-day. Alongside of these dynamic workflows, we also saw deep research, which is a workflow function. So, if you do /deepress research and you say what you want to research, it automatically will invoke one of those dynamic workflows, which is pretty cool. We also have git work trees, which is a super important feature, especially if you are working on code bases with other people or you're working on a bunch of different features within your app or your product or whatever codebase you're working in. basically lets you run isolated copies so that they never collide and then you can merge it all back together at the end. And then we also have ultra review which is another slash command you can do where you come in here, you type ultra review as you can see which will spin up sort of like a code review in the cloud. I do wish though you could type it in and it would give you that sort of like color thing the way that you do with like ultra think or some of these other things. Even if like you do workflows, it does the whole rainbow color. But either way, it's still a great feature. Some of the other things I threw in here was like the interactive charts. If you're on cloud chat and you ask for diagrams or interactive elements, it creates those right inside, which is awesome. I also love the whole recap feature where if we're in cloud code, you can see if you've been idle for a while, it basically shows you what this session has been working on. So if you've got a bunch of tabs open like I do a lot, you get a recap down at the bottom which shows you exactly what was taking place in that terminal. So for example, here I can see that we were building a slide deck for cloud code hooks, which is a video that I'll be working on and coming soon. threw in some other stuff here too, like cloud for Microsoft 365, cloud and Chrome, computer use, other stuff like that. But let's go ahead and move on to the A tier. These are all pretty much honorable mentions that I had a little bit of a hard time deciding what actually made it into the top 12. But starting with number one honorable mention, which really is in my top 12, but it's not actually like a cloud feature. So anyways, I just had to throw it on, and that's the Google Workspace CLI. So like everything that I'm doing pretty much either lives in like ClickUp or my Google environment. So, Sheets, Docs, Drive, Gmail, Calendar. So, the fact that my agent can use the CLI to navigate through everything here and I can build skills around the CLI, it can just touch things way easier and it's just very easy to maintain. So, I love that CLI. We've also got dispatch, which is kind of like the whole channels thing, and it was one of the things that Claude dropped when OpenClaw was just going nuts. And so dispatch allows you to just, you know, continue the conversation and talk to your claude from your phone specifically when you need to run like your co-work. I also threw in here the Claude desktop app which I do use a lot and it's nice to be able to be in the desktop app specifically for routines because I can manage my cloud and my local routines right from this app which is pretty nice. And I really like the ability to be able to work in a bunch of different projects and a bunch of different sessions and visually see that on this lefth hand side a little bit better than the way that I typically work in VS Code. So when I'm working, I usually have VS Code open here with a couple different tabs open or windows open. And then I usually have my Cloud Desktop app open on this other side because also when you're having a conversation with Cloud Code in something like the desktop app or even the VS Code extension, you can copy and paste the text way better because it doesn't wrap weird in the terminal. Whereas, if I'm in the terminal and I try to like copy this, it just has weird spacing. So, sometimes for those use cases, it's better to just open up the extension like this or to use the claw desktop app. I also had to throw cloud design in here because as a team, we do use cloud design quite a bit because we have our different design systems for different types of landing pages or different types of offers and it's just a really nice place. We can all leave comments, we can all look at things, and it spins up PC's really quick. Auto mode. I like using auto mode as a permission level in my cloud code. Ultra code hooks, effort levels, agent view, I really like, which is kind of like that dashboard to manage and see all the different terminals you've got running inside of the cloud code terminal, which is great. Slash, by the way, you're able to ask questions while you're not interrupting whatever your main terminal is actually doing slashcontext. I had to throw prompt caching in here because it saves us a lot of tokens, but it's just something that runs on the background that you don't even really notice. So, anyways, that is what I had for the A tier. So now let's get into my top 12. These are the features that I actually use all the time and that I really love. Let's start with number 12, which is slashgoal. So slashgoal basically lets you tell Claude, "Hey, I want you to do this and you're not going to stop until you hit that condition." So a really, really simple example, hey Claude, go ahead and do some research on the best riddles and don't stop until you can pull back three riddles that you would consider very hard to solve. So in this case, the objective that it's going for is finding those three riddles. The only thing you've noticed here is that very hard to solve is kind of subjective. So typically when you're using a slash goal, you want to have as much of an objective definition of done as you can. But this is really nice. And the thing about this slashgoal is when this came out, I think a lot of people were trying to do the demos where it's like, "Wow, my agent's been doing slash goal for 24 hours or 48 hours." And yes, that's very cool, but I've never done that and I don't really need to. What I like about SLG goal is I can shoot off a goal. I can give it a very clear definition of done and I can go to bed or I can go make lunch and I can have pretty high confidence that when I come back it actually will be done because it has verification in there. You can see this goal took 43 seconds and the output isn't anything phenomenal, but it just gives me a little bit of confidence. I can say / goal execute these skills for me and don't stop until X Y and Z. So just because you're doing SLGO doesn't mean you have to have it be hours and hours of running. It just means that you have to be clear with your prompt and set an objective criteria. And if that's a metric that you want to optimize for, that's great. For example, on our website, A Automation Society, when I refreshed this, the globe is basically an instant load. But when I first built this, the globe took like a second every time to load in. So I said, "Hey / goal, look at the globe element and don't stop optimizing this page until the user would refresh and the globe appears instantly." And now it's basically instant. And that honestly took I think it was like an hour and a half of running and trying different things until it was able to get it to this point. So that's just a good example of something that you can use / goal for. Okay, moving on to number 11. We have ultra plan. So this offloads planning to the cloud with a bunch of different planning agents and I love using this ultra plan thing when I need to really really put some effort into a big task and I want to plan first. And this one's cool because you can just type ultra plan and it does the whole little rainbow color. But let's say I want to ultra plan. Help me figure out how I can make five times the monthly recurring revenue compared to the previous month. And at this point, that's a super super ambiguous thing, right? And it says, do you want to run in the cloud? I say yes. It's going to give me a link and it will take it to the cloud. So I can click on this link right here, open this up. And then what happens is after it clones the repo and it does all of this, you know, the compute in the cloud, we're able to just basically bring that back into the terminal. or we could start executing upon the plan right here in cloud on the web. You can see back in the terminal, it's showing me that there are five background tasks that are running. And right here, we're able to actually keep working while the cloud version is still planning. Okay, so number 10, we have slashinsights. So what I'm able to do is come into here, I can type a slashinsights and it generates a report on the past 30 days of how we've been using cloud code and it gives you insights on, hey, here's what you've done a lot. Here's what you're doing bad. Here's maybe some features that you should utilize more. Here's maybe some skills you should build. So, when it gives you that HTML report, this is what it looks like. You can see it went over 1500 messages across 153 sessions, and these are the dates that it ran it from. This would be a month, but I got a new PC here, so it only was able to look through um, you know, the past 2 to 3 weeksish of local data. But anyways, we get to see what's working, what's hindering you, quick wins to try, and ambitious workflows. And we have this kind of like table of contents where you can jump around to where things go wrong or, you know, different features and things like that. So it's really cool to just check in on this every once in a while and see how the way that you use cloud code evolves. Okay, so number nine we have automemory. So I think when this released it was called autodream, but basically this is just the idea that claude is autoimproving the memory without you prompting to do stuff. So every, you know, interval of this many sessions or this many conversations or this much time, it basically just searches through what you guys did and it updates itself. It basically just dreams and makes it smarter. And yeah, by the way, I canceled all of the ultra plan. Anyways, if you come in here and you do memory, you can see automemory is turned on. So, if that's not turned on for you and you want it, go turn it on. And if you want to read more about how that works, you can go to the docs and read up about auto memory. All right, so number eight, we have agent teams. This is one of my favorite features ever. It is token intensive, but whenever I realize, okay, my session's going to end, my 5-hour window is going to end in 30 minutes, and I have 50% of my session left. Let me spin up some agent teams and just do some fun stuff. So, what I love to do, and what you can actually see right down here with these agents that are just sitting and waiting, is I love to create panels. I love to create debate boards. I love to say, "Hey, create me an agent team and have one persona be a complete beginner, have one persona be a CEO, and I just list out all these different personas." And I have them analyze something or brainstorm something and debate. And then they basically just have rounds of debates until they all agree on something. And sometimes it's just really helpful for me to hear these different perspectives and analyze it myself. And the way that you have to do this is in your do.claude inside of your settings, you have to enable the cloud code agent teams because right now it is an experimental feature. So if you put this into your settings, then you will be able to actually use it and you can pretty much invoke the agent team creation with completely natural language. Okay. Number seven, we have slashre which lets you roll back code and your conversation to an earlier checkpoint. And when we talk about how do you save context and the the idea that like everything gets cached and they get reloaded, it's actually better for you to rewind if it gets something wrong rather than saying, "Hey, no, that was wrong. Do it this way instead." And you literally can just do a slre and then you can choose the exact prompt, the exact point in this conversation on where you need to actually do it. So that is a super nifty feature. All right. So number six, we have sub agents. Background sub agents, they can get invoked automatically. You can ask it to create sub aents. The difference between the agent teams and the sub aents are that agent teams they can talk to each other. They can debate. The sub aents all work in parallel. So they can't talk to each other. They only talk to the main terminal session that invoked them. And what else is cool about the sub aents is right in here in my doclaude. Not only do we have like skills and rules and workflows, but we also have agents. So you can create these markdown files of agents that get invoked the same way a skill does pretty much. And you're customizing what that agent does and what tools it may use and any skills that live inside of an agent. So, I use sub agents a lot and it's one of my favorite features. Number five, we have /loop. This is basically the ability to just have a prompt in your terminal repeat. So, I could come in here and I could do a / loop and say every minute just remind me to use the restroom. And then when I kick that off, it basically will use a cronreate tool and it will set up that local cron to remind me to use the restroom. And the cool thing about the loop, you can see there it uses the cron create tool, is that they're they're basically bound by that session. So if I opened up a different terminal, I could have different crons over there. And if I close out of this one, it would kill that cron. But you also don't have to do / loop to invoke it. You can just say, hey, set up a recurring loop to do this every 10 minutes, every 20 minutes, every hour. And they should run for like seven days, and then they'll automatically shut down if the terminal's been open for longer than seven days. So there you go. It just got kicked off again and now I got two reminders. All right, so number four, we have remote control. The ability to drive your local cloud code session from your phone or the web. And I love this one. So I come in here and I can just like let's say I'm working on this project and I want to go take a walk or go down to the gym or whatever it is. I can just go ahead and do a remote control command and I can say enable remote control and then it gives me this link which if I open that up from my phone or I even just go into the cloud app on my phone, go to code and then I hit it up top, I can control it. So, I'm not sure if you guys will be able to see this, but on my phone, you can basically see that it gives me those bathroom reminders, right? Like, hey, um, time to use the restroom. And I can say thank you on my phone after it's running another one of those loops. But if I shoot off thank you, it comes off in my computer as well. So, this is basically completely synced locally from my phone to what's going on there. So, I love doing this, like I said, when I just want to go take a quick walk, but I want to keep that session going and keep brainstorming or keep building things. All right, so number three, we have routines. This is the ability to schedule things and not just scheduling like a deterministic Python script. You're scheduling an actual agent. You're scheduling your cloud code terminal the exact same way that you would talk to it right here. And you can literally just do like a slash schedule in order to create some sort of uh routine. But that's why I like to use the desktop app is because I just like to create them in here. And this is where I can view them. I can view it on a calendar. I can view it based on a list. I can set up a local routine or I can set up a cloud routine. And it's just really cool because you get like an actual agent on a routine rather than just getting sort of like AI calls or API calls. All right, so number two, we have the status line. This is probably one of the most slept on things ever, and it's what you guys notice down here at the bottom of my screen. I have Opus 4.8 X high. I can visually see how much of my context window has been eaten up. I can then see 27% 274,000 tokens out of a million tokens. This is one of the most important things that you need when you're working with cloud code. And it's the main reason I switched from the VS Code extension to the terminal. But now I like the terminal for more than just that reason. But it's the easiest fix ever. You literally can just do / status line as you see. And you can just say exactly what you want. You could give your cloud code a screenshot of this and say, "Hey, I want this to be shown on my status line." It's so nice to be able to see the model, the effort level, and obviously the context. So I know when I need to run my session handoff skill and when I need to clear and when I need to compact and stuff like that. And two of the features we talked about earlier, one of them was slashcontext. So we can visualize what's actually gone into this window and what might be silently eating up tokens. And then the other one was slash usage which lets us actually see um you know our session and our week and some other stats like what models we use and how many tokens and just gives us some visibility into some of this stuff. And then last but not least, number one, a lot of you guys will probably guess exactly what this is and this is skills. Like this has just been the thing that has given me so much more leverage when I'm using my cloud code. I have obviously so many skills. I refine them every day. I keep building more out and they are just the best thing ever. And the cool thing about skills is not only can you build your own, but you can pull in a bunch of skills from other people. You can get plugins that have skills and hooks and MCP servers and tons of things in them already. Or you can just grab skills because all the skills are are markdown files. They live in yourcloud. Here are the skills. You can see I've got a ton of them in here and I can share them with my team and I can continue to build these and make them better over time. A skill is basically just a recipe. It's basically when you open up your cookbook, if you want to make a lasagna, you don't always just do that off of memory. You just open up the lasagna recipe and you just follow the instructions. And so now that we've given our agent a bunch of these recipes, it knows exactly how I like to do Excal presentations or to make thumbnails or to trim videos or whatever it is. It just has to read the skill and it does it way more consistently than if you had to give it the same prompt every time. Skills can be big. They can have a bunch of steps. They can have a bunch of tools. They can even route to other skills so you chain them together. Or they can be really simple. It could be as simple as the session handoff skill, which is literally it just invokes a prompt. It's like a four-s sentence prompt that I found myself saying a lot. And now I have just turned that into one quick slash command and it's just super easy. So they're very versatile. They don't have to be complex, but they can be. And it is, like I said, just the number one feature that I'm using on the day-to-day inside of Claude because it's not only Cloud Code, but you can also use them in Cloud Chat or Claude Co-work. And that is going to do it for this one. That is our whole board of all of these top 12 features as well as my A tier, Btier, Ctier, and Dier of Claude features. Like I said at the beginning, you guys are not going to agree with this, but hopefully you enjoyed seeing the way that I was thinking about it. And hopefully you maybe learned a few new features that you might not have known existed, or maybe I've inspired you to use some of these top 12 more if you're not utilizing them too much. So, if you guys want to review the slide deck, I will attach this whole thing in my free school community. The link for that is down in the description. All you have to do is join this community. You'll come in here, click on classroom, hit all YouTube resources, and then you'll be able to see all of the resource guides, skills, GitHub repos, anything that I've ever dropped in here, completely free. But that is going to do it for today. So, if you guys enjoyed or you learned something new, please give a like. It helps me out a ton. And as always, I appreciate you guys making it to the end of the video, and I will see you all in the next one. Thanks guys.