How To 10x Your Notes: Obsidian + Claude AI Agents

Use code YOUTUBE to get an extra 20% off my AI courses here:
https://www.jointakeoff.com/

x post: https://x.com/mckaywrigley/status/194...
blog post: https://mckaywrigley.substack.com/p/c...

obsidian: https://obsidian.md/
claude code: https://docs.anthropic.com/en/docs/cl...
wispr flow: https://wisprflow.ai/
cursor: https://cursor.com/

Follow me on...
Twitter/X: https://x.com/mckaywrigley
Newsletter: https://mckaywrigley.substack.com/
GitHub: https://github.com/mckaywrigley

0:00 - intro
0:55 - setup instructions
2:40 - vault creation
5:28 - ai rules
8:00 - core chat
14:36 - stt
17:35 - cursor
21:25 - plan mode
27:12 - custom commands
34:30 - automated tags/links
40:52 - subagents
44:11 - mcp
51:28 - cloud usage via github

0:00 Hi everybody, my name is McKay and in this video I'm going to teach you how to 10X your notes with AI agents. So the 0:05 goal of this video is we are going to teach you how to use claude code which is a new AI coding tool from Anthropic 0:10 and we are going to extrapolate its coding capabilities so that we can use something that I like to refer to as claude agent which is we are going to 0:17 teach you how to build flexible agentic workflows that you can apply and all sorts of different tasks that you experience day-to-day. We're going to 0:23 have timestamps on everything. This is going to be on X. This is going to be on YouTube. I also have an accompanying blog post on my Substack. Uh, so if you 0:30 like this video, share it with friends, subscribe to my YouTube, uh, follow me on X, share the Substack post. It goes a 0:35 long way uh, to help me out. I'm just going to cover a few uh, four to five minutes of like intro. Uh, you may just want to skip that, which is totally 0:41 cool. Uh, and then we'll actually dive into those 10 workflow examples. Uh, and then we're going to close so that hopefully by the end of this, you are 0:48 completely agent-pilled and you feel like you know how to use the cloud agentic experience to build workflows 0:54 that can help you do anything. The rest of this video assumes two things. So, these are going to be steps where you 0:59 just need to pause the video and you can either, of course, just watch me do everything and then do this, or you can follow along. If you want to follow 1:04 along, you need to make sure you have Obsidian downloaded. This is a free download. You can go to obsidian.md 1:10 and you can get your free desktop version of Obsidian. Very easy to get started. So, make sure you do that if 1:15 you want to follow along. And then you will also need to make sure that you have Claude Code installed on your computer so that we can take advantage 1:21 of all of the agentic workflows that we're about to demonstrate here. So, you can go to docs.anthropic.com. 1:27 It's very easy to get to the cloud code docs and what you want to do is go to the cloud code tab here and you just want to go to overview here and you're 1:33 going to find this command. You just need to run this command in a terminal on your computer. If you're somebody 1:38 who's not technical, right, you might have seen this video. Oh, taking notes. Let me figure this out. And then be like, oh great, it's a developer telling 1:43 me how to run terminal commands. Uh don't be intimidated. This is a very easy setup. Uh you can use your AI chat 1:49 of choice. Uh if you, you know, if you're a chat GBT user, if you're a cloud user, if you're a Gemini user, uh just copy this entire page. Enthropic 1:56 makes it very easy to do. So you just copy that page, paste that into the AI app that you like to use and ask how do 2:01 I do this? It's very easy. You just need to open the terminal up on your computer, copy this command here, and install it. Okay, super super easy. The 2:08 one note, the only thing you will need to pay for of course is you do need either a cloud plan from Anthropic or 2:14 you need a cloud API key. Cloud plans start for as low as $20 a month. Uh, so if you do want to actually follow along, 2:21 get started, and for some reason you don't have a cloud plan, you will need to go to Anthropic and actually sign up 2:26 uh up here on the top right. So the two things that you're going to need to do before following along, again, you're more than welcome to just watch this, 2:32 but you need to get Obsidian downloaded and you need to make sure you get Cloud Code installed with a corresponding 2:38 Cloud plan. Then you're going to be all set to follow along. Before we get on to our workflows, we're first going to create a new vault in Obsidian. So you 2:44 can see I just have a new window of Obsidian here. And you can see I have the create new vault. We're going to click create and we're going to create a 2:51 demo vault because what I'm going to do is a first uh the first couple of these workflows are going to be things that I 2:56 recommend that you do when you get it set up, particularly this first one. So I'm just going to call this claude code 3:03 demo. I'm just going to put my vault on my desktop here, but you can put this anywhere. We're going to go ahead and create a new vault. We're going to see 3:09 the Obsidian app open here. We're going to go full screen here so we can kind of explore a little bit. So the core 3:15 experience of Obsidian is text files. So you'll see over here on the left we have our file explorer. So I could go ahead 3:22 rightclick create a new note McKay right we can say hi everyone watching. Okay so 3:27 all of these files here are just text files. So if I actually just open up this McKay file in finder here you're 3:33 going to see this is just a markdown file. Okay. So I can do whatever I want with this file. I can move this to 3:38 another device. I can use this in a different app. I have complete ownership of this file because is it is just a 3:43 text file in markdown form. You'll also notice on the right we have the graph view. So some of you are going to be 3:49 into this kind of thing, some of you are not. Uh I do think that knowledge bases get really fun over time. Right? My core 3:54 obsidian vault that you saw in that preview pane a few seconds ago. That was all years and years and years in in in 4:01 kind of the making. And my graph view is very structured. So I have this knowledge base with all sorts of connections. And the great news is with 4:06 something like Cloud Code, you can actually automate quite a few of the linking and tagging files, which we'll 4:13 we'll get to in the workflow section. Uh, but one of the things that you're going to want to do as well is you're going to want to make sure, of course, that you did get Claude Code installed. 4:21 And one of the things that we're going to want to do here is you can see we have our vault in the bottom left. So, it's really nice because I can quickly 4:26 switch between different vaults. Uh, I pretty much just operate out of this main vault. But what we can do is we can actually rightclick this and reveal and 4:32 find her. Uh, and what we can do is we can actually rightclick. Again, I'm on Mac. Some of you, of course, are going to be on Windows or Linux, but I'm just 4:38 kind of operating um with Mac here within those constraints. So, just do whatever is the corresponding function 4:44 on your uh particular operating system, but I'm just going to click new terminal at folder. Okay, so I just rightclicked 4:49 over my vault, new terminal at folder. Okay, let me just drag in my terminal here. Okay, so we're going to put our 4:55 terminal over here on the left side of the screen, and we're going to keep Obsidian over on the right side. What we can do is we can just run Claude on our 5:01 vault. And now we have a full instance of cloud code working on our vault. Now I can say something like hi what's in 5:10 the McKay file. Just verify that cloud does in fact work. We click on our file 5:15 explorer. We'll see we have a McKay file. You can see the cloud code is now intelligently searching. Okay. You can see it found the McKay.md file here. And 5:23 you can see that it said hi everyone is watching. All right. So at this point we're all good to go. Now let's jump into the workflows. Workflow number one 5:28 is going to be about Claude's rule system. So the very first thing that you're going to want to do once you have cloud code running is you're going to 5:35 want to type /nit. And you're going to see this init command pop up. You're going to see that it initializes a new 5:40 cloud.md file with codebase documentation. So that's highlighted. We're just going to hit return. And 5:46 what's going to happen here is cloud code is going to intelligently create an initial claude.md file. You can see 5:52 claude allcaps.mmd. This is going to be the rules file that is essentially injected as the system prompt to every 5:59 single request that you make with cloud code inside of your vault. Okay, so this pertains to every single folder, every 6:05 single file. It's basically going to be the very very top level prompt. So if you have anything, you can see cloud 6:10 code is asking me for permission to do things here. I'm going to go ahead and say yes, you have permission. This is 6:15 another one of those things you'll start to learn over time is the agentic nature of claude is very powerful, but it's 6:21 also very controlled. And you can see it's now asking me for permission to create a new file. So if I accept yes 6:27 here, you're gonna see I have three options. Yes, yes, and don't ask again this session. No, and tell Claude what to do differently. So if I selected 6:33 three, I would get an opportunity to kind of give it some additional text, kind of prompt it in a different 6:38 direction. In this case, this looks pretty good. And we're just going to kind of roll with it just for demo purposes. In the case of your actual 6:44 personal vault, you're going to want to create this file. You're going to want to kind of go in here and you really do want to actually take the time to 6:51 customize this to your liking because this file is going to have a significant significant impact on the quality of 6:57 responses, how Cloud actually kind of uses your codebase or your not your codebase. I'm so used to working in 7:02 code, how it works with your notes uh inside of your vault. So, you can make this as specific and personalized as you 7:09 like. Here, we're just going to kind of roll with what it gave us out of the box here. uh we can kind of assume because it's a fresh vault and we did in fact 7:16 tell cloud it was a fresh vault that these are probably going to be pretty good and we'll come back to this more 7:21 later as we try some of the other workflows because we'll show you how you can kind of use those workflows in conjunction with updating this rules 7:27 file to get better for for uh performance. But anyways, the thing that you want to do at the very start of your 7:33 cloud code experience in Obsidian is you just want to make sure that you use that slashinit command. You want to let 7:39 claude create this claude.md file at the very base root of your project folder of 7:45 your vault so that these rules. Okay. So if I actually exit cloud code here uh clear my terminal here. If we were to 7:52 initiate a new session, what actually happens is now this prompt gets injected into this new session. So this now 7:57 applies to every single session which is pretty cool. Workflow number two here. We're going to boot up a new session of 8:03 cloud code. Again, just as a reminder, now everything that is in our claude.md 8:08 file here is going to be injected inside of our prompt. So, Claude is now going to follow these instructions. Uh, just 8:14 want to keep uh one of the things you're going to notice as we go from workflow to workflow. It's a lot of them build on top of each other. So, just want to 8:20 bring that up one more time here. But, we're working in what is called Cloud Codes interactive mode. So, this is very 8:25 chatbased. You know, you kind of saw me working with it in the last workflow section, but this is a very natural kind 8:32 of UX paradigm, right? Text in, text out. You just ask the agent what you need and it does it, right? Very simple 8:38 and it just kind of operates this entire thing inside of the terminal, which is just a little bit different, you know, than the web browser or the app chat GBT 8:45 or Claude or Gemini experience that you're already familiar with. It's just this time we're doing it with cloud code as a command line tool inside of the 8:51 terminal here. So let's just go over a couple of kind of the core workflows in standard chat. A little bit more of a 8:57 boring section here, but this stuff is important. So if I just say, hey, pick a random topic, right? We're just chatting 9:03 with cloud code naturally like we would in the chat GBT interface for example. And in this case, we're going to have 9:09 chat or cloud code rather pick an interesting topic here, the Fibonacci sequence. And now I can say awesome. 9:17 Please create a new note about that. What's going to happen here is Cloud Code is now going to write a new file to 9:25 our Obsidian vault here and it's actually going to create this file. It's writing this. You can see it's actually 9:30 updating the tokens in real time. So, it is actually kind of writing this under the hood. You're just not seeing it 9:35 stream in like many of you are probably familiar with. And then you can see it's asking us for permission to create the 9:41 file. Again, the great thing is if I wanted to allow it to always create files, I could go into my rules file and 9:46 I could actually allow it to do that. In this case, we're just going to accept this manually. And now you can see we have a Fibonacci in nature file here, 9:54 right here inside of our vault. So, one of the great things about Obsidian, Obsidian, because it is all markdown 9:59 files. AI models are super super good at markdown. So, you're always going to get really nicely formatted files, it's 10:05 always going to have a really good idea of how uh to kind of construct things. If we go to read view, we can kind of 10:10 see it in a little bit nicer preview here um like you would normally. Then we can kind of go back to edit mode and 10:16 edit this file. Uh but one of the things you'll notice like at the bottom here you'll see we have tags. This is how you get interesting things like 10:23 knowledgebased automation. So one of the things you'll notice in our cloud rules file file um cloud rules file here is we 10:30 have things like tags. So if we wanted it to never have tags, we could do something like you know do not use tags, 10:36 right? And now our rules file is updated and it wouldn't use tags anymore. So that's like a really lightweight example 10:42 of how you could customize this rules file to make sure that Claude Code is writing notes in the style in which you actually wanted to write that. We're 10:48 going to undo that for now because we do just kind of want to demo some of the different things you can do. But now 10:53 Claude Code has kind of written almost this like basic Wikipedia entry style um article here uh inside of our Obsidian 11:00 vault here. And we can do things like, you know, see this welcome file. We actually don't need the welcome file 11:07 anymore. And what you'll notice is Claude is going to ask me for permission to remove it. Here you can see it's 11:13 actually searching for us to find the welcome file, right? And you can see how Claude's search capabilities can get 11:18 really interesting because imagine, you know, something like my own personal vault that has like thousands of files at this point. It can actually just 11:24 dynamically search that entire thing without you having to tag it. Here or, you know, have to kind of manually 11:30 provide a file path. You can see it's asking for me for permission. It found the file. We're going to say yes. And 11:35 you're going to see that file disappear. Okay, pretty cool stuff. Now, if we wanted to tag a specific file, say we 11:41 wanted to go back into that Fibonacci, we can actually just type in at, we can kind of start typing. We're going to see the Fibonacci here. I can hit tab. 11:47 That's just going to autocomplete that. And we're going to say, um, can you add 11:53 an interesting fact section at the bottom, but above the tags? And now 12:00 we're going to see it actually go and edit the specific file that we tagged. Okay, so we tagged that Fibonacci file. 12:06 But you'll see it will in fact ask us for permission for the edits. Again, I'm going to keep repeating myself here for a little bit and then I'll stop. Uh, but 12:11 you are going to want to approve the file. You can see all of these green lines here. Those are files that are being added. If you ever see red here, 12:18 those file, those lines, those characters, those are getting removed. So, we're going to accept that. You can see we now have an interesting fact 12:24 section that is right above the tags like we asked for. Now, say we wanted to clear our message history. Say we 12:30 started working on a new task or something. What we can do is we can just type /cle. I'm going to tab to autocomplete that. Hit return and this 12:36 is going to clear my session and boom I am now ready for a new chat. Okay, so 12:41 few things. Okay, cloud code can create files, it can edit files, it can remove files, it can also create folders. Okay, 12:46 so if we wanted something like create a math folder and put the related files in 12:53 it. One of the things you'll notice is we actually cleared our me message history. So it doesn't actually have any memory about the Fibonacci in nature. 13:01 Okay. Uh so what's going to happen is it's actually going to realize oh Fibonacci sequence kind of related to 13:06 mathematics here. So what we're going to do is we're going to create a new folder and we're going to put that file in it. And it's intelligent enough to 13:12 understand that it needs to do that. And not only is it intelligent enough to do that in this setting where we just have literally two files in here, but it can 13:18 do that as your vault grows. So this is again the great thing about an agentic system is cloud code is really just an 13:24 agent here that's really good at knowing when in and how to use tools. In this case, it's searching for all sorts of 13:30 different mathematically related operations here. It's doing this agentically. We could actually hit CT 13:35 controllr and that will actually expand. If I looks like I need to hit control E, you can see a lot more uh insight into 13:41 what's going on here. We're just going to toggle that off with control R. And we're kind of now back to our familiar 13:46 format here. And you can see it's now trying to run make deer uh which is going to make directory. Okay, it's 13:52 going to create that math folder which we accept. And then it's going to do one more operation here to put that file inside of that folder. Okay. And boom. 13:58 Now our Fibonacci and nature file is inside of that folder. And we did that dynamically without having to specify that this is the file we wanted. Right? 14:05 Cloud code is very smart. It's kind of able to figure out what you need. The goal is you just want to prompt as clearly uh and kind of directly as 14:12 possible. So very cool powerful agentic system. Again, workflow 2 is really just about some of the core basic chat 14:18 features. We just wanted to show you a few demos of this so that you understand uh kind of what's going on and just uh 14:24 make sure that you're not intimidated by this and all the different UI things that are popping up. It's just an agent and you just talk to it like you would 14:30 any other AI app and it kind of just autonomously does things for you. You just need to kind of approve the operations and customize the system to 14:36 your liking. In workflow number three, we really want to emphasize the importance of using a speechto text tool. So this one's kind of optional, 14:43 but I promise this one is going to change your life. So this is the speechto text tool that I use. It's 14:48 called Whisper Flow. The product specifically now is called Flow. You can see uh the URL up here. But all you do 14:54 here is you're just trying to increase basically the amount of tokens that you can output to the AI. So this is just 14:59 going to speed up your workflow. You can speak tokens. You can speak words about four times faster than you can type 15:05 them. So if you're somebody like me who's basically working with AIS all day, every day these days, uh it's just 15:10 much more efficient. If you're using a speech to text tool, I can just kind of dictate my own thoughts directly to the AI without having to hammer on a 15:16 keyboard. So, we're going to skip installation here. There's a little bit of a setup here. Uh, you don't even 15:22 necessarily have to use this tool in particular. May maybe your native device has speech to text functionality. Uh, 15:28 maybe you have a different tool. The importance is just that you use a speech to text tool and I just wanted to kind of highlight the one that I use. So, the 15:34 reason I use Flow here is because a uh it has a really nice Mac and a really nice iOS app. So, as kind of an Apple 15:42 guy, as you're probably catching on to here, uh, it's very easy for me to kind of sync my whispers or flows or whatever 15:48 the I actually don't know what the unit is they officially call them here, but one of the things that's common for me is like maybe I'll go on a walk, I'll 15:55 record the, you know, use the mobile app to to record some of my thoughts. Then I'll come back on desktop and I can just 16:01 like paste things into notes and then have the AI organize it and things like that. Uh, but in this case, what we'll do is we'll just jump back to our 16:07 Obsidian on the right, cloud code on the left setup here. And what I'm going to do is I just hold down my function key on Mac OS. Again, the reason I love Flow 16:13 is just because user experience is really great. This isn't like sponsored or anything. Um, uh, if anybody from 16:18 Flow is watching this, maybe we can get you guys some free credits or something. Um, but I'm just going to hold down my function key and then I'm going to talk. 16:26 All right, Claude, I want you to create a new note. It should be your three 16:31 favorite dinosaurs and why. Okay, we're just going to send this off. And you can see that was very quick. It 16:37 was very instant. All I had to do is hold down my function key and speak. And this isn't the greatest demo here. Uh 16:42 this is particularly useful when you just kind of want to ramble for a while. Uh so you can see we just accept that. We now have three favorite dinosaurs 16:49 from cloud here. U but one of the things especially in like my coding workflows is I'll have these sessions where I'm 16:54 like rambling about a feature I want built for like three or four minutes. And I can just hold my function key. I don't have to type. I don't have to 17:00 really like think through it. I can just dictate to the AI and cloud code is powered by smart enough models that kind 17:05 of figure out what you need even if the speech is like a little messy or jumbled. Um so very very nice way to 17:10 just sort of 4x your output. Uh this is one of those things that once I see people get hooked or people try this 17:17 they get hooked immediately. This is a total game changer to not just your workflow with this particular workflow but anytime that you're actually 17:23 interacting with an AI especially if you coders out there are working with cloud code in actual code format. Super 17:29 unbelievable tool. Definitely recommend you try a speechtoext tool to increase the amount of tokens you can output. 17:35 Workflow number four. So, you're probably wondering, wait, I thought we were using cloud code. Why is cursor here? So, we're going to talk about how 17:40 you can actually use cursors tab feature, which can be really, really powerful here. Um, if you want to kind 17:46 of get a little crazy here and use a code editor as your notes editor. So, let's go show you kind of how to do that 17:52 and I'll explain how this can actually make sense in practice. One thing that's worth noting just really quick before we 17:57 do that is that this is probably the most optional of everything that we talk about here. So, I know the speech to 18:02 text was also optional. I think speech to text will totally change your life. I think this one's a little bit more niche. This is something that I do like 18:08 to use, especially when I'm very frequently iterating on a specific file inside of my vault. Um, some of you who 18:15 are non-coders may not want to work in a code editor, which is totally understandable. But nevertheless, we kind of want to show you uh why cursor 18:22 tab can be useful for something like Obsidian. So, we're going to go back over to our Obsidian here. And what I'm going to do is I'm actually just going 18:28 to rightclick on my vault. We're going to reveal this in Finder. Now, this is the important thing about Obsidian, right? Is all of these text files you 18:34 own. They're on your computer. You can use these in any app you want. So, say I was working in this three favorite 18:39 dinosaurs or something. Uh, what I could do is you have two options here. You can open the individual file up inside of 18:45 cursor and work like that, or you can open the entire vault inside of cursor. In this case, I'm just going to open up the specific file that I want to use. 18:51 So, I'm actually just going to drag this under cursor. Again, I'm kind of assuming that you have a cursor open 18:56 here. So, I'm just going to drag this file down here. You're going to see we have our three favorite dinosaurs MD. 19:02 Okay. Again, one of the great things about markdown files, they translate very well to code editors. So, cursor 19:07 does in fact make sense. Now, you can use this file with all of Cursor's own agent features if you wanted to. Uh 19:12 there's nothing stopping you from doing that. I particularly find that cursor tab can be very useful. So, cursor tab 19:18 is cursor's native autocomplete feature. So, one of the things you'll notice is if I change this to four, this is going 19:23 to start using cursor's tab model. So, if we actually just do return, I don't even have to do return. This is how great cursor tab is. You can see it's 19:30 actually just populating a new dinosaur here. And now I can just hit tab, hence the name of the feature here. And that's 19:35 just going to autocomplete my file. Now, if I save this, you'll see we have a little dot. If I save this and drag this 19:41 window out, we go back into our Obsidian, you'll notice that this file is saved. Why is that? Because this is 19:46 just a single text file. It doesn't matter where we have it open on our computer. You could even build your own user interface for all I care to open up 19:53 markdown files. The point is that every single one of these text files you own. You edit it one place, it gets edited 19:59 the other place. Everything syncs well. So if you're somebody who wants to like do a lot of editing in a file, right, 20:04 cursor tab can not only do creates, but it can do edits, it can do deletes. So if you're working in a really long note, right, say we wanted this to be bullet 20:10 points, for example, instead of having to go and make every single one of these a bullet point, it's smart enough to kind of figure out what I'm doing. And 20:16 now we just can kind of like tab through this. Boom. I didn't have to do that manually. Now I can save it. I could of 20:21 course just ask clot code to do that same thing for me. So if we drag this out and kind of open our terminal here 20:27 back open, uh we can say please convert and we'll actually tag that file. Okay, 20:32 we're starting to tag we're starting to kind of pile these features on top of each other which starts to get where the 20:38 workflows get really interesting, right? This become a power user. Even use speech to text here because why not? Please convert the bullet points to 20:44 numbers. And boom, we'll send this off the clock code and it's going to be intelligent enough to go switch these 20:50 bullet points back to numbers. Uh, so this is just an example really, really lightweight of using cursor tab, but I 20:55 think if you start to use your imagination, you can see where that gets interesting. Again, very very optional here. I know a lot of people are like, 21:01 well, I don't want to, you know, pay for all these different products. I want to kind of consolidate. So, no need to do that, but a lot of people who are 21:07 already using cloud code also happen to have cursor plans. If you're somebody like me, uh you can even if you want 21:12 open your entire vault, open inside of cursor, open up a terminal tab inside of cursor, use cloud code inside of cursor 21:18 while using tab, right? You can really get pretty wild with some of these workflows, but we just wanted to show you a little bit of a simple version of 21:23 that. So, let's move on to the next one. In workflow number five, we're going to talk about plan mode. So, plan mode is accessible inside of cloud code if you 21:29 do shift tab once. Shift tab is going to do auto accept edits on. So, let's 21:35 actually just do a quick little side bonus tip here. Autoaccept edits on accepts everything that Claude does. So, you've seen me kind of approve Claude's 21:41 actions. If you don't want to manually prove that, you can turn auto accept edits on and it's just going to accept everything. Now, this can obviously be a 21:48 little bit dangerous. Um, I recommend not turning this on until you've kind of coordinated the permission system inside 21:54 of Claude on your own a little bit or at least just gotten familiar with how it works so that you can kind of keep an eye on it. Uh, because, you know, you 22:00 don't want it to do things like editing your entire uh, you know, obsidian vault with all these notes you've diligently kept, right? Uh, so you want to be a 22:06 little bit careful with that one. So, one of the nice things that you can do to help it stay a little bit more on track though is if you hit shift tab one 22:12 more time, you're going to see plan mode turns on here kind of in green here. At least with the theme that I have running 22:17 here. What plan mode does is it actually makes sure that cloud code doesn't do anything. It doesn't perform any 22:23 actions. It's just going to take your prompt, create a plan, and then you have an opportunity to either approve that 22:28 plan or reject that plan. Let's go ahead and take advantage of our speechto text tool here and give it a little bit of a 22:34 basic prompt and then we can kind of see how plan mode works. I want you to go to the anthropic docs and I want you to 22:39 figure out how the cloud code SDK works. I'm interested in building an express server to build uh automations with the 22:47 cloud code SDK. Okay, so you can see that is now inside 22:52 of my prompt here. I send this off to cloud code and because we're in plan mode, cloud is going to kind of think through this and reason throughout 22:58 things. One thing that you'll notice too here is we actually have uh access to web search. So Claude can search the 23:05 web, which is kind of a nice bonus little workflow tip here is we're going to actually take advantage of number two 23:10 here. So anytime now Claude will not ask permission for going to the docs on 23:17 anthropic.com. Here you can see we get a nice URL preview. So now, because I've selected that option, it can just autonomously do that without having to 23:23 ask me for permission every time, which is particularly useful in this research case because it's probably going to go 23:28 through a different uh few different URLs here to figure out what's kind of going on with this SDK. So, couple bonus 23:35 tips here already, right? We've gotten the auto accept mode uh and then we also have web search abilities. So, cloud 23:40 code does have quite a bit of uh feature depth to it if you really really want to get crazy here. But, we can see it's 23:47 just trying to figure out what's going on. It's not writing any files, right? It's just doing research. So, note-taking, of course, is a very broad 23:54 discipline. I kind of batch research in with note-taking. You know, one of the things that I do when I'm writing code 23:59 is I do a lot of research. I like to take a lot of notes in my projects. So, inside of my vault, you know, I have a folder for my projects and then if I 24:05 need Cloud to go like search some docs or something, figure out how like a certain API works or a certain software 24:11 library works, I can just hand that right off to Cloud Code. Can go do that research task autonomously while I'm working on something else. I can have 24:17 plan mode on so that I can kind of come back and make sure that all of the things that it writes directly to my vault. I kind of approve all that sort 24:24 of stuff. So, let me just kind of sit here for a second and I'll speed this up once it's done with its research task and we'll take a look at plan mode. 24:30 Looks like cloud code is done with my plan. I can kind of scroll up. You can see it's boxed in here which is sort of the user interface inside of the 24:36 terminal. Anthropic's really done a quite a good job with trying to make the user interface inside of the terminal good despite it just being a terminal 24:43 based program here. So, you can see we're given two options. Yes, which this is just going to accept the plan and no, 24:48 which is keep planning. And what I can actually do is I can give it a little bit more feedback here. So what's going to happen here is we should just kind of 24:54 read through this. So what we're going to do is we can see okay cloud code SDK allows you to run cloud code as a subprocess. Okay, everything here looks 25:01 pretty good. So what I'm going to do is I'm going to hit note and keep planning. What we're actually going to do is we're going to say uh we want this documented 25:09 in a cloud code SDK note. So, we're just going to make sure it knows, hey, we actually want to take all 25:16 the research that you just did. We actually want to add it to Obsidian Bault. Of course, it's not going to add that yet because plan mode is still on 25:22 as indicated below the text input here. And you can see we would now like it to proceed. So, let's go ahead and hit yes. 25:28 And one of the things you'll notice is it actually turns auto accept mode on by default. So, we can actually shift to 25:33 tab back to regular here. You can see Cloud Code is actually tracking its own internal to-do system, which is kind of 25:39 cool. So, what's going to happen is it's probably going to ask me for permission to create the note. I'm actually just going to go back to autoac accept edits 25:45 on just to show you how that works here because it's kind of one of the little bonus tips we're going to do in this workflow number five here. So, we can 25:52 just kind of let this thing work for a little bit here. Um, I might speed this up just a little bit so you guys aren't watching forever. So, as you can see 25:58 here, Cloud Code is now auto accepting the file creation here. It wrote to that file. It's kind of tracking its internal 26:04 to-do system. You can see over here in our vault, if I click on cloud code SDK, we now have this entire documented um 26:11 kind of explanation of how to use the cloud code SDK with an express server. You can see we even get things like code 26:16 blocks. Again, cloud code very very good at using markdown like all AI models are these days. You can see it is now done 26:22 with the task. We were able to kind of figure out what we needed from it. You can do this with much more complex tasks 26:29 of course. So if you want to use plan mode for something that has like a dozen plus steps, totally can. You can just 26:34 continue to iterate on the plan back and forth with cloud code until it is configured to your liking. Hit that yes 26:39 accept plan button and then you can either auto accept everything it does or you can kind of improve it manually one 26:46 step at a time. Obviously in this case we did auto accept edits but that's plan mode. That is a way to kind of plan complex tasks. Really comes in handy 26:52 with things like researching um things about like organizational tasks where you really want to make sure cloud is 26:57 going to be doing the right thing. Ton of opportunity here. Again, one of the things that's great about Cloud Code is just how flexible it is, right? 27:03 Obviously, it's powerful, but it's also very malleable and flexible. You pretty much get it to do whatever you want. And one of the best ways to get it to do 27:09 whatever you want is to use plan mode. Workflow number six, we're going to talk about custom command. So, this is very, 27:15 very powerful. A custom command inside of Cloud Code is basically like a reusable prompt that can be very detailed and it can kind of outline to 27:22 cloud code what task you want done. So, if you have things that you're doing very often inside of your notes, this 27:28 can be something that's very powerful. So, for example, in my own workflow, one of the things that I do is I have kind of a daily template. And one of the 27:34 things that I do now with cloud code is instead of having to go like copy my daily template and like paste it in and 27:39 configure it with some kind of starter information based on the day, I just use the daily template, which of course 27:45 isn't here because it's not in this vault. And then what I'll do is I'll just kind of use my speech to text tool. Talk about some of my goals like, hey, I 27:51 want to get XYZ done today, right? Use speech to text. And now I kind of combine those two things into a reusable 27:58 command. This daily template command and that goes it creates my new daily template. I have some instructions on like where I want that organized, how I 28:04 want that formatted, all that kind of stuff. So what we're going to do in this section is we're going to show you how you can actually create one of those on 28:09 your own. This is going to get a little bit weird and I apologize for that just because of a little bit of a jank in how 28:15 Obsidian works. So, if I actually drag in my finder here, again, the way that you can access your Obsidian Vault uh 28:22 folder kind of at the very parent is you can just click the right click rather in the bottom left of Obsidian here and 28:28 just reveal and find her. That's going to pop up in your vault. So, you just want to make sure you go in your vault. And one of the things that Claude does 28:34 is it actually creates a claude folder. So, if I do uh shift command dot here on 28:39 Mac, this is going to show hidden folders. the convention, at least on Mac, I'm actually not sure how it is on 28:44 Windows, but if you have folders that are kind of uh prefixed with a dot, it will actually hide them in the finder by 28:51 default. So, this can get a little bit tricky for things like Git, for things like GitHub folders, which is very 28:57 popular if you're a coder. Obsidian, in this case, has aobsidian folder that it hides with a bunch of different 29:03 configuration stuff. And then claude, of course, has the Claude folder. So, you're actually going to need to edit 29:09 this outside of Obsidian, which I know is like a little bit janky, not the best workflow, but hey, what can you do? 29:15 Because in Obsidian in the file explorer, it hides all these dot folders just like Mac does by default, but 29:20 there's no way to show them. There's a little bit of an interesting design decision, but that's a conversation for another day. So, what we want to do 29:26 inside of the cloud folder is we're actually going to create a commands folder. So, what we want to do here is 29:32 we're going to create a new folder called commands. And I actually just need to drag that inside of Cloud here. 29:38 So what your file structure should look like is you should have your vault. Okay, so we have the entire vault here. We should have thecloud folder, which at 29:45 this point you should now have. Uh Claude should have created this for you by now, but if you don't have it, you need to create a cloud folder. And then 29:51 inside of thatcloud folder, you need a commands folder. So this is where we're going to be creating our new commands. 29:57 So what I'm actually going to do is I'm actually just going to drag this folder open inside of cursor. We're going to be 30:03 using cursor as our text editor. So you can use again you can use any text editor that you want. Uh you can see we 30:09 just have this commands folder open. So we are just going to be creating new text files. And all commands are also 30:15 markdown files. So if we do something like um get time get-time markdown we could 30:23 just create a heading get time and we'll just say please please get me the current time. Right? 30:32 Okay. And then what we could do here, you'll see uh if we go back to our uh finder here, you'll see we have that 30:38 file of course. And now what we can do is we can do slash. But wait, we actually have to restart our cloud 30:43 session. So this is something I see that people do all the time is they'll create a new command. And you actually have to exit out of your session, start a new 30:50 one, and then if we do slash, we actually now get all of our custom commands, a slash command. So now you'll see that that get time command is in 30:56 fact loaded. So what we can do is we can just start typing get. You can see it's already highlighted. So, we can just tab autocomplete that and send this off. And 31:03 it's going to go ahead and get the current time. So, you can see it's Monday, July 7th. Um, I'm trying to get 31:09 a couple extra sneak sneak a couple extra workflow videos in here before I post this tomorrow. Um, but uh so you 31:16 can see that's like the most very basic version of a command that's obviously not particularly interesting. So, what 31:21 I'll show you is I'll show you uh something that I use, right, which is like that daily template command. So, I can do daily- template markdown. Again, 31:28 all these have to be markdown files inside of the commands folder which is inside of that.cloud folder. And the 31:34 reason that we're having to edit this inside of cursor again you can use any app you want is because obsidian hides those. So we can't actually directly uh 31:40 do that in obsidian. So for daily template I would do like daily template and you know maybe we want uh this um 31:48 little you know we'll just accept whatever cursor tabs uh going to give us today. And what I can do here is I can 31:54 also at the very start say you are given the following context. This is something 31:59 I do all the time. Claude actually has a way to give the context that you give to a command after you use the slash. So if 32:07 I say hi there, right? If we want this hi there injected into this command. What we actually have to do is we 32:13 actually have to use this dollar sign arguments and all caps here. We want to make sure 32:18 we spell that correctly, of course. And what's going to happen is that hi there would get injected right here. So what's 32:23 very common in my own personal commands, right, is I'll give it a heading of whatever the name is and then I'll say you are given the following context. 32:29 I'll include that dollar sign arguments keyword here and then I'll actually go about what I need. So you can do all 32:35 sorts of things in here, right? You can say um you know one of the things I could do is I could say like please 32:40 create this in the and we'll call this um you know daily- updates folder. 32:48 We could save this uh and then we could have this like basic template. So what's going to happen now is of course because I created a new command we do need to 32:55 reset our cloud session here and now I can do /daily template and now I can 33:00 type here I can speak here with my speechtoext tool and I could just say hey I'm super blocked by the last few 33:08 lessons of my cloud code obsidian workflow. 33:13 And what's going to happen is we can send this off. This is going to uh actually get injected into my prompt 33:20 here, right? You can just think of commands as prompts. And what's going to happen is it's going to create a new file inside of daily updates. And you'll 33:26 even notice our vault doesn't have daily updates. That's totally fine because Cloud can do that all in one operation. And now we have this new daily updates 33:33 folder here. Uh and let's say, you know, we wanted to go look at that. You can see those blockers are now injected 33:38 right here. We got the title. Uh we got kind of the daily update for January 8th. And we could be as specific as we want, right? you'll see like this isn't 33:45 a very clean date format. So, one of the things that we would do inside of our command, right, is we would just specify, hey, this is how we want the 33:51 date. This is how we want things formatted. This is where we want that file created, right? Maybe, you know, we haven't talked about MCP yet, but maybe 33:56 we have some MCP connections. And we can say like, hey, uh, you know, pull from this MCP tool or use this tool inside of 34:02 that connection, right? You can get super crazy with commands. Commands are incredibly composable and powerful. I'm 34:07 just showing you the very, very, you know, we're just scratching the surface here. So, commands are, think of them as 34:12 like reusable prompts. the kind of recipes that you can give to cloud code for things that you're doing all the time, right? In my own personal coding 34:19 workflow, right? I have commands for things like generating plans, things for using git operations, all sorts of 34:24 stuff. So, that's one of kind of the most powerful workflows you can really go crazy on, and I highly recommend you 34:29 experiment with it. In workflow 7 here, we're going to get into things like automated tags, linking, and 34:35 organization. So, this is a really, really powerful one, especially for those of you who are super into things like knowledge bases. If you really like 34:41 things like wiki style links and tags in your notes, if you want to kind of make more connections across your different 34:47 files, especially as your vault grows, I know that's something I really like because, you know, as your vault grows and grows and grows, can be difficult to 34:54 kind of maintain that knowledge graph. Uh, which, you know, I'm somebody who can definitely get a little bit obsessive over my knowledge graph. I 35:00 just think it's the compounding effects of it over years and years is really, really cool. But one of the things that you can use cloud code for is to help 35:06 you do a lot of that more automatically. So, I'm going to show you an example of this for tags, but you can also 35:12 extrapolate this for wiki style links and for file organization in general. 35:18 So, one of the things that we're going to do, and I highly recommend that you do this because this kind of sets the foundation of your system, is you want 35:24 to instruct Claude Code to create a tags file for your tagging system. So what 35:31 we're going to do here is we're going to create a little bit of a prompt to cloud code and we're just going to kind of work through what a really really basic 35:36 v1 version of that would look like. So I'm just going to say please create a and I do this in all caps tags.mmd file 35:44 at the root. This just means at the very base level of the vault. In the file you need to define rules for our tagging 35:53 system with hashtags inside of our obsidian vault. 36:00 Okay. And what's going to happen here is Claude at the very base of our vault is going to create this tags file. This 36:06 tags file is very nice because now anytime that we want Claude code to autonomously handle tagging in any of 36:12 our files, in any of our folders at any time, we can always tell it to kind of reference our tags.md file. And this is 36:20 something that we want to keep updated. Uh not only can we keep it updated, but Cloud Code can keep it updated. So we're 36:26 just going to kind of create this right away. Okay. And you can see we now have this tags file. So when you are creating 36:32 this for your own system, you're really going to want to put in the time to actually curate this, of course, according to how you like tags done. In 36:38 this case, we're just going to kind of accept the defaults, but this is very similar to the kind of claw rules in 36:44 that tags is almost kind of the grounding layer of our tagging system. So you can see it came up with a bunch 36:50 of different things. Um, you obviously again customize this to your liking. But one of the things that you want to do is 36:55 you want to now tell cloud code that in its kind of cloud file like hey anytime you go about doing tagging make sure you 37:02 reference this tags file because this tags file is kind of the source of truth. So I'm going to use speech to text here and just do this autonomously 37:08 with cloud code. Okay. Now we need to make sure to update our claude.md file 37:14 here. That's the rules for your system prompt. We just need to make sure that anytime you go about tagging, you 37:19 reference tags. We want this to be the source of truth for our tagging system. And we want to make sure that you also 37:25 keep this updated over time. I'm just going to send that to cloud code. And what's going to happen now is 37:31 it's going to actually dynamically alter this cloud. MD file here. So you can see it's creating a very basic little to-do 37:38 system here. It's going to ask us permission to edit the file. It's going to add somewhere in here some notes on 37:43 how it should actually go about using the tags file. Again, in your own workflows, you're going to want to make sure um 37:50 that it uh does that autonomously here. So, you can see here we got one update here, which is this tags line right 37:56 here. You can see see tagging system section below. And now it's actually creating the tagging section. So, we're going to accept that. And now down here, 38:02 you can see important always reference tags.md for tagging rules. So, this is where clouds in it system prompt now, 38:08 right? This is the basically the rules for cla code is it's going to reference that. It's going to see this every 38:13 single time we run any cloud code session. So anytime it updates our tags, it's now going to go about that. So 38:20 let's actually ask us for a demonstration. Please give us a quick demonstration to make sure that this works 38:25 and we should hopefully see it whether it edits an existing file or creates a new one. So it looks like it's going to 38:31 check tags.md for appropriate tags and then create a sample note here. Again, this this makes a lot more sense once 38:37 you have your own personal vault with many many different files here. Um, worth noting too, you can actually just 38:42 create vaults for different things. I know people who will go about it that way, too, if you want to kind of silo things and have, you know, certain 38:47 pieces of your, uh, workflow be more separate. Uh, you don't necessarily have to give Cloud Code root access to every 38:54 single thing. You can kind of divvy it up, which is kind of cool. So, we have sample meeting notes here. Just go ahead and click sample meeting notes. And 39:00 presumably, you can see there's actually an error with how it must do numbers in tags. So, we would maybe have to go 39:05 update our tags roles to account for something like that. Kind of teach the AI how to work. But overall, uh, got a 39:11 few things done here, which is nice. Presumably, this follows the system rules here. We're not going to like dive 39:16 into that. Again, I think you get the idea here. This is this is where you want to define your rules for your tagging system. And you just want to 39:22 make sure inside of your main rules file, your cloud. MD file, that you kind of mention the tagging system so that at 39:27 the very minimum just knows to reference that file and to keep things uh up to date over time. So, that's kind of a 39:33 really nice way you can create an automated tagging system. You can do things like combining this with commands, right? So, say I wanted to 39:39 create a command for tagging a file. Maybe I have a command called like tag-file, right? And maybe in the inside 39:44 of that command, I say like, hey, uh, you know, go inside of whatever file I've tagged and update it with some 39:51 tags, you know, and you have some instructions there. Maybe we do tag folder. Maybe this is for specifically 39:56 tagging folders, right? You can kind of start to chain commands with dynamic organization and things like that. And 40:02 then, of course, you can do things like creating a wiki link uh wiki wiki links. 40:08 I always see a little bit of a tongue twister there that right to do something like this uh if you wanted wiki style links which is very popular in 40:14 knowledgebased applications. So you could do something like that where you you know create a new note called like wiki links 40:21 d obviously you don't do the d in title you get what I mean right if you want things for uh organization this is 40:27 another one I use in my own personal vault organization and then anytime uh cloud code creates new files and folders 40:32 it kind of has a good sense of structure in which I like things organized. So, there's a couple different ways you can build automated systems around 40:38 organizational features, but in particular, uh, tags, very core one that's very nice to do. The wiki links 40:44 one, very nice to do, and then organization of like files and folders, another really nice one to do. So, highly recommend you take advantage of 40:50 that in conjunction with things like slash commands. Workflow number eight. So, this is going to be a quick one. Uh, surprises me. A lot of beginners in 40:56 particular don't know about this, but you can actually ask Claude Code to use sub aents. So, it will actually be able 41:02 to do things in parallel. Uh so you don't have to wait for one task after another to be done. This is particularly 41:08 useful for things like researching tasks. Uh but you can kind of use your imagination. So really quickly I just 41:14 want to show you uh we're going to do sub agents and we're also going to do a bonus of showing you how to use deep 41:20 thinking which is kind of another little bonus tip here. But to use sub agents in cloud cloud code you just ask for them. 41:26 So one of the things that we want to do is let's just say we want u let's actually just dictate this to the AI and 41:31 we'll show you uh what's going on here. I want you to spin three sub aents up 41:36 for a new research task. We are going to create a new note that is called AI model pricing. I want you to look up the 41:43 pricing of OpenAI models, cloud code models, and Google models. And I want you to put those in a table inside of 41:49 that note. Um, and I want you to price everything per million tokens. 41:55 Okay. So now we can send this off. Again, very speedy if you're using speech to text software. You can now see 42:00 it's going to create three sub aents. Just like we said, it's creating a to-do system. So presumably what's going to happen is you're going to see task pop 42:06 up for these three things. And these are all going to be happening at the same time under the hood. So this is also going to take advantage of things like 42:12 web search under the hood. Very cool. So we're starting to chain things. One of the things that I'll do in a lot of different versions of my projects is 42:18 I'll have a slash command for sub aents where I kind of have different sub aent tasks for things like research, for 42:24 things like writing notes, all sorts of different stuff. Uh so what we're going to do here is we're going to let each one of these run in parallel. Okay? 42:30 Okay, so instead of having to wait for this first task and then asking for the next one, we're just doing this all at the same time. Okay, so sub aents are 42:35 very cool. You can see it's actually using web search for each one of these. So I'm going to go ahead and speed this up and show you the end result here. 42:42 Looks like these tasks just finished up. We should now be getting on to the note creation point. All I had to do in some 42:48 of these tasks was just accept a few web search tool requests here. Obviously, all of that could be configured to run 42:53 autonomously, especially if you did accept autoedits here. Uh, which we're going to do for the actual generation of 42:58 the file here so that we don't have to accept that. So, in just a few seconds here, we should hopefully see that table. Okay, you'll remember we did 43:04 specifically ask for a table so we can see how well the AI does. Again, because it knows how to use markdown, it's going 43:10 to be very reliable. Uh, really nice little use case here. So, obviously kind of a silly little example here, but you 43:16 can see how this can get kind of nice. It can. You can see here it's now writing to AI model pricing. Looks like 43:21 we do in fact have that new note and it did create the table. It's a little squished here because of my window size. But if we actually go uh to read view 43:28 here and let's actually just go full screen you can see we have the OpenAI Anthropic and Google models. You can see 43:34 they're all priced out. It dynamically found this based on its own web search. So nice. Oh, looks like it even did some 43:40 key insights and pricing notes which is kind of cool. Some notes on prompt caching batch API. Okay, very useful stuff. So, one of the ways that I use 43:46 this in my own workflows, you know, I'll do a lot of uh research on things like code libraries. I apologize, some of this is a little bit code heavy on the 43:53 examples here, probably some of you more casual people who aren't coders are like, h, that's kind of annoying, but uh 43:58 again, it's personal to me. You can personalize your cloud code experience to you. This is very flexible. Um, but 44:03 that's an example of how you can use sub agents. You just ask cloud coder for the sub agents and it's able to kind of spin up these parallelized tasks to do a 44:10 bunch of different things at the same time. Workflow number nine, we're going to show you how to use MCP servers with your note-taking experience using cloud 44:17 code. So, this is where things can get particularly powerful. And if you actually read my corresponding blog 44:22 post, you can find the links for that below. I wrote a post called Claude agent where one of my big points and 44:28 reasons for doing this video is that Claude code is so much more than just claent 44:34 agentic system that's like super good at knowing how and when to use tools, which is why I kind of like to call it Claude 44:39 agent. uh because you can build agentic workflows for any type of task, any type of job. Uh especially with things like 44:46 MCP servers, which if you're not familiar with MCP, this stands for model context protocol. This is kind of a new 44:51 way in which people are building tools and integrations that can very easily fit within AI models and some of the 44:57 most popular AI tools that you know and love. So, one example that I'm going to be showing you here is something called 45:03 context 7. This is something that I particularly use in a lot of my research tasks. Uh again I mentioned earlier in 45:09 this video and you guys are going to love it. Another code adjacent example here uh hence the life of a developer 45:16 for you here. Um but one of the things that this MCP server is really useful for is getting up-to-date information 45:21 about software libraries and documentation. So you can see if we go to context7.com 45:27 is worth noting guys you can use MCP servers for all sorts of stuff. There's MCP servers for things like Google Drive, right? So, say I wanted to pull 45:34 info in from my Google Drive into my own personal vault. I could actually just connect to the Google Drive uh MCP 45:40 server, get that set up, and then boom, I would be able to do slash and you would actually see the MCP server for 45:46 like Google Drive, right? And then you'd be like, "Hey, can you actually pull this file inside of my Google Drive?" It would go find that, create a new note 45:52 for it, right? Maybe you want to pull four or five different files. But the point is like all these external tools, right? Maybe you're somebody who's running a business and you you want to 45:58 use the Stripe MCP to query customer data or something. Uh maybe you have like a customer call coming up. Maybe 46:04 you need like the billing information. I don't know. Just coming up with random things on the fly here for you. Point is 46:10 MCP is very extensible. There's all sorts of MCP servers popping up these days. I'm just showing you one very 46:15 particular example and something that I actually use uh most days. So you can see they have this MCP uh tab at the 46:20 top. We're going to go ahead and click that. We're going to scroll down and we're going to get some instructions on how to actually configure this. So, if 46:26 we scroll down, you can see they have a little bit of installation section here in the readme. And one of the things you'll notice is they have install and 46:33 cloud code. So, in our case here, we're going to use the remote server connection. It's worth noting, right, if 46:39 you just want to search a popular tool that you use in Google, right? You know, XYZ MCP, it's going to be pretty easy to 46:45 find the documentation for whatever server you're looking for. There's so so many these days. And side note, not only 46:51 can you use preconfigured ones, pre-built ones by existing companies and people all over the world, you can also 46:57 create your own, which is how you can get really, really crazy with cloud code, especially when you're taking notes. You can kind of build your own 47:03 workflows. A very quick example I'll give of this is I actually have in my own vault, I have a customuilt MCP 47:08 server where I can input a YouTube link. I can use a custom prompt command and it will download the transcript of that 47:15 YouTube link, take notes on it, and put it inside of my vault. Okay, so you can get pretty crazy with this. I'm just showing you a very quick example. Um, 47:21 but in this case, we're going to copy this command. And then what we need to do is we're actually going to need to 47:26 run this in our terminal, not inside of cloud code. So, one thing you can actually do in cloud code is you can turn bash mode on with this exclamation 47:33 point here. Um, you can see uh exclamation point for bash mode. This is now going to effectively run as a 47:38 terminal inside of the text input of cloud code. So, I don't have to exit out of my session to run terminal commands 47:44 here. And so, I'm going to paste this in. This is going to add this MCP server here. So what I can do now is I can do 47:50 /mcp and you can see we are now connected to the context 7 MCP server. 47:55 So you'll see we have the second MCP server here that we're connected to. This is actually the deep wiki MCP server from Devon. Uh this is from 48:02 Cognition Labs. This is another tool that I use their MCP server for a ton for coding research when I'm working on 48:08 uh different projects. Um but in this case we're going to go ahead and use Context 7. So we're going to reset our chat here and we're going to ask a 48:15 question. We're going to say use context 7 to find how to do generate object 48:24 with enthropic models inside of the versel 48:29 SDK. So this is actually going to ask us for permission for us. This is very similar to how claude uses its other tools. You can see resolve library. So 48:36 the first thing that context 7 is going to do is actually going to pick which documentation to pull from. Okay, so we're going to approve that. It's going 48:42 to search for Verscell AI SDK and then it's actually going to go ahead and dynamically find that inside of its 48:48 libraries. If we go back to Context 7 here, you can see Verscell AI SDK. If we search for it on their site, presumably 48:53 it's pulling from this one right here. You can see we got a little terminal popup. That means our message is ready inside of Cloud Code. And now it's 48:59 actually going to dynamically search for it instead of me having to go in the UI and I can just offload that task to cloud code. So we're going to approve 49:05 that here. You can see there are some parameters here. So what it's going to do is it's basically going to pull the 49:11 cached docs that context 7 has so that I can really really quickly get an answer to my question here. Okay. So it looks 49:17 like we got a little bit of documentation here on how to use that library. And again we did this via the context 7 MCP server which is really 49:24 cool. So now I can say awesome create a new note in code slash ideas 49:33 chatbot. What this is going to do is going to create a code folder. This will create an idea folder and then this will create an AI chatbot folder here. So 49:40 what we can do is we can actually just turn auto accept on with shift tab. This is going to go ahead and allow cloud 49:46 code to autonomously handle the creation of this without us needing to approve this which is very cool. So you're starting to see the pieces of how all of 49:52 these things are starting to come together and form these really powerful workflows. So you can see there's the code folder, there's the idea folder, 49:57 there's the AI chatbot. Looks like it did create that as a folder. So we'd maybe want to be like, hey, actually 50:03 create that as a note. It's probably going to put a note in here anyway. Uh, so I'm not going to worry about it. That's something you could very easily 50:09 iterate on with cloud code. Um, but again, MCP servers super powerful. You can build your own. There's so many on 50:15 the internet for all sorts of things, right? Say you're somebody who, you know, I'm also somebody who uses notion a little bit for forming personal wikis 50:22 that I want to be a little bit more presentable and maybe share with friends. I could connect to the notion MCP server and then I could pull things 50:27 from my notion into here. or I could actually pull things out from my personal vault and maybe I want to create a new notion page with info 50:33 that's from my Obsidian vault. Right? So, not only can you kind of read information and bring stuff in, but you 50:39 can also take information from your Obsidian vault and you can use that with outside sources. Okay? So, this thing is 50:44 very cool. Uh MCP is a very very deep rabbit hole that I highly recommend you dive down when pairing it with taking 50:51 notes can get super crazy. One of the last things I'm just going to ramble on for like 20 seconds and then I prom 50:56 promise we'll move on to the next thing is you can start to really build these agentic workflows. Again, this is why I 51:01 call it cloud agent and not cloud code because say you're somebody doing sales, right? You hook this up to a few MCP 51:07 servers and suddenly you're almost like doing your job in conjunction with cloud code instead of the terminal and maybe 51:12 you know maybe you're not even using Obsidian, maybe you have your own, like I said earlier like a custom UI built out. There's just so many possibilities 51:18 here. are really just scratching the surface of what AI agents can do. And all of this is just so so exciting to 51:23 me. And of course, the use case we're trying to use to demonstrate all of this is taking notes. This is going to be the 51:28 last workflow demonstration. Workflow number 10. This one's going to get a little bit weird. Uh, and this is really 51:33 meant to just kind of demonstrate something more broadly, which is that you can actually deploy Cloud Code in the cloud and have it do things 51:40 autonomously when you're on the go. You can do things like building WhatsApp integrations. Uh, you can build your own 51:45 custom app to manage this. All sorts of things to do. The easiest way to demonstrate this and to get started with at least just experimenting with this 51:52 type of workflow is going to be with Claude's custom GitHub action. So what I'm going to do here is I'm actually 51:57 going to use Cloud Code itself to create me a new Git repository. And what we're going to do is we're going to set up Claude Codes GitHub action on our GitHub 52:04 repository. So GitHub also an interesting thing to integrate with an Obsidian vault because it also gives you 52:10 version control on your vault. So this is something uh that I'll experiment with from time to time on kind of test 52:16 vaults. I actually don't necessarily do this workflow with GitHub on my main vault. I actually have my own 52:21 infrastructure built out to handle that these days. But we're just going to demonstrate this to you so you can get a feel for it and just see kind of the 52:28 magic of a workflow like this. So what we'll do is we'll say I need you to initialize a new GitHub repo and push it 52:33 to my GitHub as a private repository. Okay, so I can actually just tell Cloud 52:38 Code to do this. Cloud Code is intelligent enough to figure out how to use the GitHub CLI in conjunction with 52:44 initializing a new Git repository on my machine here for this vault. So, I'm probably losing some of you 52:50 non-technical people if you're not like a software engineer or developer or something like that. It's going to be a little bit weird for you. Uh you may 52:56 just want to sit tight and watch this one. Uh but it is pretty cool. I promise the payoff on this is cool. We're going 53:01 to have cloud code run get. This is going to initialize a new empty git repository here. And what we're going to 53:06 do is it's going to commit our files to GitHub and create a new repository with the GitHub CLI tool. Okay, so a lot of 53:13 different things happening. You can see those of you who are developers can see, okay, it's adding all files and then it's committing with initial commit 53:18 message. This just means that all the files that I have in my vault are getting added to a new repository that we are about to push to GitHub 53:26 cla Okay, so I'm just giving it a name. You can see cloud code's intelligent enough 53:32 to know it needs to ask me for a good name. So, it's going to create a new private git repository called Cloud Code Vault on my GitHub here. Going to push 53:38 this. Again, GH is just the command for the GitHub uh CLI, right? Just like 53:43 Cloud Code is a CLI. GitHub also has a CLI and this is what it's using. So, we're going to go ahead and proceed. I'm 53:49 going to let that create and then I'm going to go open that up inside of GitHub. You can see we're on github.com. 53:55 This is in my new vault that I have on Git. Okay. So, the idea here is you can do a couple things. If you want to use 54:02 version control in your vault, very cool thing to do. You can kind of get a running history of everything in your vault. Obviously, if we were configuring 54:08 this for real, we would do things like adding the store to get ignore and all that kind of stuff, but we're just kind of we're just kind of going through the 54:13 motions here. Now, the important thing is inside of cloud code, what we can now do is we can do slashinstall github app. 54:19 You can see this is a predetermined uh or kind of a pre-existing command inside of cloud code. This isn't like a command 54:24 I came up with. This is on everybody's instance of cloud code. And what we can do is we can just tab that. And then we're going to hit return. And this is 54:31 actually going to use the repository that we just set up and it's going to install the GitHub app on a repository. 54:37 So in my case on my account, Cloud is allowed to install the GitHub app on any repository. You're going to need to 54:42 install that for the first time or configure what repositories this particular GitHub app has uh access to. 54:48 Uh I can kind of just skip that because it already has access. So I can just kind of close that. And now you're going 54:54 to see if we go back to cloud code. We just need to initiate um this kind of 54:59 initial uh thing here. So you can see we can now select GitHub workflows to install. Both of these are checked. Uh 55:06 we're actually going to just accept both of these. So we'll just hit enter. And then you can see create a long live token with your cloud subscription. So 55:12 you can either use your own API key or your own cloud subscription token. We're just going to go ahead and use our 55:18 existing uh cloud token here. So we can just sign in directly with Anthrobic. Again, we've gone over this, but to use 55:24 cloud code, you do need either an API key or a Cloud subscription. In my case, I'm just using the Cloud Mac 55:31 subscription, the $200 one because I'm such a power user. You can see we now have a very introductory PR. This has 55:36 some information about how this particular integration works. We're just kind of speeding through this because you actually have to approve this before 55:42 the Apple work. You actually have to create that and then merge it and then you're going to be all set. Okay, so 55:48 that's merged. Uh we can delete that branch. Everything is ready to go. So now, how does this get interesting? 55:55 Let's switch over to the phone. I apologize guys, the audio is going to get a little bit weird here. You can see I am now on the GitHub app on my phone. 56:01 Okay, so let's imagine for a second that I'm on the go. Let's say I'm on the walk and I actually want Claude to do some research for me in the background in the 56:08 cloud. And then when I come back to my computer, what I can do is I can actually go see its research inside of my vault. So what I'm going to do is I'm 56:13 going to go into issues here. And what I'm going to do is I'm going to create a new one in the top right here. And I'm just going to say test integration. 56:21 We're just going to do something really basic here. And I'm just going to say please make a new file in the vault 56:28 in the slash test folder called hello 56:33 world and write a funny joke. Not a runny joke, a funny joke 56:40 for my viewers. Okay. So what we're going to do is we're going to submit this. This is going to create a new 56:46 issue inside of my GitHub repository here. And what I can do is I can add a comment. You'll see this comment button 56:51 at the bottom. And because we've installed the Claude app, I can now actually tag Claude. You don't need to see uh anything there. And what we can 56:58 do is we can say, "Please handle this. Thanks." We'll do a smiley face. We'll 57:04 go ahead and leave that comment. And what's going to happen is this is actually going to kick off our GitHub 57:09 action. So anytime that we tag cloud code inside of issues, inside of PRs, it's actually going to go autonomously 57:15 handle this, which is super super cool. So we actually have to wait just a little bit to get set up. We're now 57:21 going to switch back over to the computer here so that I can show you how this works. So you guys can see this is 57:26 now triggered from my phone. Again, I apologize for the phone audio when we're back to the computer audio. You can see 57:33 theoretically, and I know I'm not literally doing this right now, but theoretically, I could have just been completely on the go, uh, not having to 57:39 be around my computer at all. And this cloud integration is now just running in the cloud, and it's actually running this, you know, without me needing to 57:45 accept anything. Um, all sorts of stuff. So, there's a lot of ways you can configure this to be a little bit more personalized to your vault. And again, 57:52 at this point in my own personal vault, I have kind of my own server that's running where I can trigger these 57:57 things. But I think the GitHub action is a really really quick way to at least get started and start playing around with this type of workflow. So what's 58:03 going to happen is cloud is basically going to look through all the comments in this issue. You can see it's going to check if a folder exists. You can see 58:08 it's going to create a hello world. You can also go in your GitHub. There's the actions here. You can see my issues 58:14 actually running is out of this tab. You can see it actually just finished up. So if we go to issues here, click test 58:19 integration. You can see uh it actually did in fact create that new file. Hello world. So what we can do now is we can 58:25 go back to our vault and because this was actually uh created. It was actually committed. So we go to code here. You 58:32 can see we're going to have uh presumably probably a new branch here. Okay. Awesome. So this is a new branch. 58:38 So what we want to do is probably just grab this branch name here uh and then actually ask Claude uh Cloud Code to 58:45 merge it. So, what we'll do is we'll go back to cloud code here. And what we're going to do is we're going to say, "Awesome. 58:51 We just had and we let's actually use speech to text here." By the way, you saw me clear that text really fast. If 58:57 you hit escape twice really quickly, it will clear your input. Okay, we just ran an automation. There's a new branch. We 59:03 need that to get merged into the master branch. And then we actually need you to pull so we get the latest version of our notes. 59:09 So, again, Cloud Code is very intelligent. It's going to know how to use the git CLI uh the github CLI rather 59:15 to go ahead and accomplish this task. So it's going to fetch all the remote branches. It's going to see the existing branches. It will ask us for permission 59:21 because we haven't configured permissions here. You can see that cloud issue two right there. It's going to go ahead and get merged into main. Okay. So 59:27 it looks like it's going to pull origin master and merge it with this branch. Awesome. Now it's going to go ahead and 59:32 allow us to see inside of test and inside of hello world. We now have a joke. Why don't scientists trust atoms? 59:37 Because they make up everything. So the workflow that we basically simulated here is that if I was on a walk, if I 59:44 was on my phone, I could go ahead and create a new issue on GitHub, I could use Cloud Code's GitHub uh integration 59:50 and basically kick one of these things off in the cloud, come back to my computer later, pull that research. So 59:55 again, silly little example here, but the goal is to get the wheels turning in your brain that you take the simplicity 1:00:01 of what I just demonstrated and you actually apply those concepts into practical real world workflow examples 1:00:06 into your own vault. Hopefully those 10 workflow tips will help you 10x your note-taking with AI agents, in 1:00:12 particular Cloud Code. The point of making this video is that Cloud Code is so much more than just a coding tool. It is a multi-purpose agent. It's very 1:00:19 flexible. It's very malleable. So, if you found this video helpful, helps me a lot. If you guys subscribe to my YouTube 1:00:24 channel or you follow me on X or if you are reading on Substack, go check out the accompanying blog post that we have 1:00:31 or that I have rather called Claude Agent. Give that a read. Um, but if there's any other things you guys would 1:00:36 like to see me do, if you want to see like an advanced version, maybe I use my own personal vault a little bit more, 1:00:42 uh, let me know. I'm super open to doing more types of videos like this. Obviously, I do a lot of videos in the realm of the coding world and things 1:00:48 like that, but definitely open to more general purpose agentic workflow videos if this is something you go like. Uh, 1:00:54 one quick note to close on too is we do on takeoff we have our cloud code lessons. So, we have 23 lessons which 1:00:59 are 2 hours and 22 minutes. We have a couple more coming out just a few more days uh on some project stuff that we've 1:01:04 been working on. Uh this is a really great way to get up to speed on all of the features of cloud code everything in 1:01:10 its toolbox. So if you are somebody inclined who is like wow cloud code seems really cool I should learn more about this. Really really excellent 1:01:15 resource we put together so that you guys can go learn how to use cloud code. So thank you for watching again. Uh share this with your friends if you 1:01:21 found this helpful. Uh goes a lot to help me out. Uh thanks for watching guys. We'll be back soon with more videos.