// transcript — 459 segments
0:00 Quick Demo
0:01 All right, so I just brain dumped a manual process into chatbt and it spit
0:04 this out for me. And now I'm going to come into Claude code and just say, "Hey
0:08 Claude, build me this end workflow. Then I'm just going to paste in that
0:10 information and shoot that off and I'll check in with you guys when we get that
0:14 workflow back." Okay, so that took about 3 minutes and you can see now that the
0:16 workflow has finished. So let me open up this link. Wow. I mean, for me to be
0:20 able to just brain dump into Cloud Code and have it spit this out. Now all I
0:23 have to do is connect my credentials and I should be good to go. And just a real
0:26 quick breakdown what actually happened here. You can see that it's not perfect.
0:30 It creates a to-do list. It starts to slowly go through each of its tasks and
0:33 check them off as done. It's searching through documentation to figure out what
0:36 nodes to use and how to set them up. And the cool thing about Cloud Code is when
0:39 it runs into issues, it literally figures out why it went wrong and then
0:42 it fixes it. Right here, you can see I see there are some issues with the code
0:45 node parameters and Slack node configuration. Let me fix these by
0:48 updating the workflow. It happened two more times. I see the Slack node
0:51 configuration uses this. Let me fix it. I see the code node needs this. Let me
0:54 fix it. And that's the power of using these super strong models like Opus 4.5
0:59 in cloud code paired with MCP servers and tools and skills. So I'm going to
1:02 show you exactly how we set this up today. And trust me, I know it might
1:05 look a little intimidating, but I'm going to make it super super simple. So
1:08 let's get into it. All right, so you guys know that my job is to make
1:10 Claude Code / VS Code Setup
1:11 complicated and intimidating things as simple as possible. So even if you're
1:14 not super interested in this specific use case of cloud code building Nin
1:17 workflows, I still would highly recommend that you watch this video. I
1:20 really think learning cloud code is the future of AI automation. I've been all
1:24 in on Naden, but lately I've been going all in on cloud code and I'm going to be
1:27 bringing a lot more of this type of content to my channel. And so I think if
1:30 you take the next hour and just follow along with this video and do the steps
1:33 that I'm showing in this tutorial, it will be an amazing investment of your
1:37 time because inevitably later when you realize you need to get into Cloud Code,
1:40 it's going to make it so much easier because we're going to break everything
1:43 down today super simply. So the first thing you're going to want to do is go
1:46 to your browser and download Visual Studio Code, which you can just go to a
1:50 browser and type it in and then it will look like this. and you'll just go ahead
1:53 and download that for your system. Visual Studio Code is just an interface
1:56 that's going to let us connect and use Cloud Code through an extension. And the
2:00 reason why I chose VS Code is because it's just the cleanest and it looks the
2:04 least cody in my opinion because you could do it on something like Cursor or
2:07 of course you could just access Cloud Code in your terminal like this. But I
2:10 don't want to be looking at this at all and I know that 80% or more of you guys
2:14 don't want to be looking at this either. So we're going to do it in VS Code and
2:16 it's going to be pretty simple and I'll show you guys how. All right. All right.
2:19 So, once we're in here, you're going to go over to the lefth hand side and
2:22 you're going to click on extensions. Once you get here, you're going to type
2:25 in cloud code and it should pop up right there. And you'll click on cloud code
2:28 and then just go ahead and install it. When you install this, it should
2:31 automatically pop up and just have you sign into your Anthropic Cloud account,
2:34 which would be the same as if you went to Google Chrome and typed in Claude and
2:38 you were talking to Claude right there. It'll have you log over there so that
2:42 you can use your plan in VS Code with the extension. Now, from here, you're
2:44 going to go over to the lefth hand side and click on explorer. And this just
2:48 basically means that we have to open up a folder to work inside of for this
2:52 specific project. And so what I did in my documents folder, I made a folder
2:56 called aentic workflows where I'm just going to have a bunch of different
2:58 projects for cloud code. And I'm going to open up this one called naden builder
3:02 which has nothing in it. So it's a fresh project. And I'm going to select it. And
3:05 what you can see here is now on the lefth hand side we have this project
3:09 which is n builder. And different files will pop up as we start making things in
3:13 here. And then on the right hand side, we've got the actual interface where
3:16 we're going to talk to Claude. And the way that you do that is on the top
3:18 right, you'll see this little extension button. And you can open up Claude Code.
3:22 Close out of this. And now we have Claude Code ready to go working in this
3:27 project for us. And this looks very similar to when you would use like a
3:30 Project “System Prompt”
3:30 chat GBT or Claude on the web. So the first thing you want to think about
3:34 doing in a project is actually telling Claude Code what you want to do in this
3:38 project. So, giving it a bit of a system prompt as far as this is the environment
3:41 you're in and this is what we're going to do together. You can even see right
3:44 here it says, "Tired of repeating yourself? Tell Claude to remember what
3:48 you've told it using claude.md." So, what we're going to do is come over to
3:51 our explorer. I'm going to click on new file and I'm going to type in claude.md.
3:56 And this just creates a markdown file for us that is essentially think of it
4:01 as the system prompt for this project. So, when we're in the file, we can
4:04 actually look at it. We could also close out of it. And if we wanted to, we could
4:07 open it back up again. But that's just like the interface that we're working in
4:10 here. Now, the first thing that I actually want to do is I want to put in
4:13 that system prompt. And what we're going to do is use Claude to build that system
4:16 prompt. One thing that I want you guys to take note of is down here we've got
4:19 different modes. So, right now we're on bypass permissions, which just means
4:22 that Claude doesn't have to ask us anything. It can just go ahead and take
4:25 action and execute things, which is really good to be in later when we want
4:28 it to actually run for us. But right now, what I want to do is go to um plan
4:33 mode. You can see there's also ask before edits, there's edit
4:36 automatically. But what we're going to do pretty much every time before we
4:38 actually say, "Okay, go ahead and do that for me." Is we're going to work on
4:41 plan mode. And if you're cycling through right now and you don't see the bypass
4:45 permissions, the reason why is because you're going to have to go to your
4:47 settings, which is in here, you're going to go ahead and type in clawed code. And
4:53 then when you get there, you should see allow dangerously skip permissions. And
4:56 you have to turn that on if you want to allow bypass permissions mode. Now, it
5:00 seems dangerous, but it's really not because you are still in control and you
5:03 can watch it and once you set up different like prompts and different,
5:06 you know, instructions, it will pretty much work inside that environment. But
5:09 that's just how you get that setting. So, I'm just going to explain to Claude
5:12 what I want to do and then it's going to help me plan out to create that claw.md
5:16 system prompt. So, I'm just going to say, "Hey, Claude, I need you to help me
5:21 create the cloud.MD file for this project." Basically, what I've set up
5:24 this project for is just to help me create workflows in NAN using you. So,
5:29 I'm going to give you access to a few different tools. One of them will be the
5:33 NAN MCP server and one of them will be Nadn skills and you'll use those in
5:39 combination with my request to help build workflows in my NAND instance that
5:44 are high quality. So the end MCP server and the end skills. Right after we set up
5:49 this file, I'll show you guys what those are and explain how those are going to
5:52 work together with us. But right now, let's just basically take a look at
5:55 what's going on. So when you send off a request to Claude in this environment,
5:59 it's going to start doing some stuff. But the cool thing about this is if you
6:03 just play around with it and talk to it, it will basically walk you through
6:06 what's going on in its mind. So you can see it started off by saying, I'll help
6:09 you create the claw.md file for this NN workflow builder project. Let me first
6:13 explore what's already in here. And you can see on this left-h hand side in our
6:16 file explorer, there's nothing in here already. So, it's going to search
6:19 through it for a bit and then it's going to come back and give us some more
6:22 information. And I know that cloud code might seem intimidating, but really it's
6:26 just as simple as if you can talk to a developer or if you can talk to a
6:29 brainstorm partner, you can use cloud code. The key is not being able to know
6:33 how to read code and understand what it's doing. The key is can you clearly
6:37 communicate what you want? Because if you can communicate the end goal, claude
6:40 will help you get there. Because what it does is it asks you questions like,
6:43 "Hey, did you mean this or this?" or "Do you want this or this?" And because Opus
6:47 models and sonnet models are so good, they'll help you do pretty much anything
6:50 you want. And if you ever get confused, you just say, "Hey, that doesn't make
6:53 sense. Can you explain that to me?" Or, "Why'd you do that?" And it really,
6:56 really helps you get things done fast. So now it's asking us some questions
7:00 already. So it says, "What ended MCP server tools will be available?" And I
7:04 will say, I will share those details. It asks about the skills and I will once
7:07 again provide those details. And then for the end setup, it asks what our
7:10 setup is. And I'm just going to say that I'm using Enident Cloud and submit those
7:14 answers, which once again, all of this information is going to help it create
7:17 its system prompt so that everything works better in the future. Okay, cool.
7:21 So, it says, "Awesome. You're using Eniden Cloud and you will provide the
7:25 details for the NIDAN MCP server and the NIDAN skills." And so, now that it's
7:28 asking for that information, I'm just going to go ahead and open up those
7:31 GitHub repositories and show you guys what they are and what they do. And then
7:34 we're going to just give Claude the URL so it can search, read, research, and
7:39 The Project Tech Stack
7:39 just figure out for itself. But of course, before we do that, you know, I
7:43 have to pull up a whiteboard. So what we're looking at is basically at a high
7:45 level, what we're actually building today in this Nadm project. So we have
7:49 us over here on the lefth hand side, and what we're doing is we're talking to
7:53 cloud code. And ultimately what we want cloud code to do is talk to our nen to
7:58 build workflows for us. Now cloud code by itself is smart and it knows JSON but
8:04 it doesn't yet understand nodes and NN expression syntax and configuration and
8:08 parameters. So what we need to do is give it access to different tools to do
8:13 so. So the first one of those is NN MCP server which is this NN MCP server right
8:17 here. It's on GitHub. I'll leave a link to this down in the description. And
8:20 also shout out to this guy. I'm not even going to try to pronounce his name
8:24 because he built both this MCP and he also built the skills which we'll show
8:28 you in a second. So in this repo you can actually read about what this is and
8:31 what it does but I'm actually just going to go to the website that he built for
8:35 it which basically says to connect your AI to NN and let it build, deploy and
8:39 debug workflows for you. So it makes it super easy. Now how does it actually do
8:42 that? Well, it has complete node coverage. So all of the 1100 nodes with
8:48 99% property accuracy. So, parameters, options, and all that kind of stuff
8:50 perfectly documented. It's got over 2,700 workflow templates, so it can look
8:54 through those, pull different elements from them, and get some inspiration, all
8:58 these other benefits, of course, and the ability to create, update, activate, and
9:02 manage your NN workflows. So, really simply, all I'm going to do is copy the
9:07 URL from this GitHub repo, go into VS Code, and first of all, just say here is
9:12 the NN MCP server, and link that in there. And then let's go back into
9:16 Chrome and let me talk about the NN skills which once again is a GitHub
9:20 repo. And basically what it says here is NN skill set for cloud code to build
9:24 flawless endm workflows. So basically what this is is it teaches the AI model.
9:28 So it'll teach claude code how to use this. And so if you've never heard of
9:32 like a claude skill, basically just think of them like system instructions.
9:35 And the benefit here is that claude only reads through them when it needs to. So
9:38 if you think about this master cloud code agent having all of this context of
9:41 all the nodes and all of the different ways to use them, every time you talk to
9:45 it, it would use so many tokens to read through all that. And so essentially
9:49 these are tools because it knows I have access to this stuff, but I'm only going
9:52 to call it when I know that I need it. And so it basically uses the skills, it
9:56 uses the MCP server, and then it's able to actually build really nice workflows
10:00 in NN. And so in the NN skills repo, if you scroll down, which you guys can read
10:04 about, here are the seven skills that it gets access to. It knows about
10:06 expression syntax which is really important. It knows how to use that MCP
10:11 server that we just looked at. It knows workflow patterns. It knows how to
10:14 validate them. It knows how the nodes are configured. And then it also knows
10:18 how to code both in JavaScript and Python. So now when we give our cloud
10:22 code access to these two things, it basically just becomes supercharged in
10:26 building NN workflows. So, I'm going to go ahead and copy the URL to the NIN
10:31 skills. Open up our cloud code and then just say here are the Nadn
10:35 skills. You can go ahead and research them so that you can make the cloud.md file
10:42 better. So, go ahead and shoot that off and it will continue to build out a plan
10:47 for our system prompt for this project. You can see what it's doing right now is
10:50 researching those and it's using a web fetch for those URLs just like we kind
10:53 of wanted it to do. Okay, so here's the plan that it came up with. It has an
10:58 overview, a research summary, it has the skills, the structure, and all this kind
11:00 of stuff. You could obviously go ahead and read through all of that, and then
11:03 you could tell it to keep planning. But in this case, let's just go ahead and
11:06 auto accept this plan so that it will update that claude file for us. Okay, so
11:12 it finished up that claude.md file and it goes over environment, MCP tools,
11:17 skills, workflow building, safety rules, all this kind of stuff. So if I actually
11:20 click into this file, we should now see that we have a pretty nice breakdown of
11:25 what we actually want to do in this project. So you can see for the workflow
11:27 building, we have to understand requirements. We search templates, we
11:32 research nodes, we build, we validate, and then we deploy. And so just giving
11:36 our project a lot more structure as far as like what files it will have, what
11:40 the end goal is, it just makes the whole thing better. because otherwise if we
11:44 didn't have a claw.md file, we would just start creating files willy-nilly
11:47 and then claude itself would have no idea how to look through those and what
11:49 Installing n8n MCP and n8n Skills
11:50 it's trying to do. Okay, but the next step now is we need to actually go ahead
11:56 and install those two things, the MCP server and the skills. So, what I'm
11:59 going to do is we're going to open up the slash, which has a bunch of
12:02 different slash commands. That's how you can change the thinking model. You can
12:04 do different plugins and things like that. What I'm going to do is slash
12:10 clear, which just lets us clear the conversation. And now we get to start
12:13 fresh. So in the GitHub repo for both of these things, if you keep scrolling
12:17 down, it'll basically go over how to install these things. And you've got
12:21 these different like mcp.json files that you could do in order to install this
12:25 stuff. Same thing if we go to the skills and you can see if I scroll down, it
12:28 shows installation. But what's so much easier is what we did earlier where we
12:33 say, "Hey Claude, here are the two GitHub repos we want to get in our
12:36 project. Just go ahead and install them." So I'm on bypass permissions mode
12:40 so I don't have to say yes anything. And I told it to install the server, install
12:44 the skills, and I also said make sure they are accessible by cloud code in
12:47 this project so that we can build flawless end workflows. And I'm going to
12:50 let this run and then I'll check in with you guys when we get some sort of
12:53 response back. Now the one thing about bypass permissions is it still won't
12:57 bypass permissions if it doesn't have enough information which is great
13:00 because in this case if you remember the end MCP server is supposed to also be
13:05 able to create manage and update workflows in your instance and I didn't
13:09 give it my instance information. So here it says do you have a cloud instance
13:11 that you want to connect for workflow management? If yes I need the URL and
13:15 API key. So I'm going to say yes I have ended in cloud and submit that. And then
13:18 it's probably going to come back and say okay cool can you give me that
13:22 information? So let me just show you real quick how we can get that. And
13:24 right on Q would it ask for both of those. So I'm just going to say I'll
13:27 provide them. So let me go over to NN and let's actually get that information.
13:32 So I'm in my NN cloud instance and first of all let's get the URL. So up up top
13:36 I'm just going to grab everything before the slashhome which is you know it should end incloud
13:41 if you're on the cloud and I'll come back in here and I'll say I have both of
13:44 these credentials ready. So it's actually just prompting us for them. So
13:48 enter your end API URL. So, I'm just going to go ahead and paste that in
13:52 there as you can see. And now we're going to head back into editen and I'm
13:54 going to go to my settings. I'm going to go to edit API. I'm going to create a
14:00 new one. Call this one cloud code demo and copy that API key right there. And
14:04 now, right on Q again, it's asking us for the key. I'll say other. I'll paste
14:08 it in right there. And then we will go ahead and submit those answers. So now
14:12 it should be able to create the MCB configuration like we see right here.
14:16 And then it's also going to clone those NN skills for us. You can see there's
14:20 the to-do list. So, it did the end MCP server. Now, it's installing the skills,
14:23 which you can see just got put in right there. So, right here we have the MCP
14:28 JSON file with our information. We have the NDN skills which are being added in
14:32 which if I click under here, we should see now we have those seven skills that
14:36 it talked about in the repo. And so hopefully Claude will be able to look at
14:39 these two things to help us actually build out our workflows after this is
14:43 done. Okay, so I know we're moving a little fast, but let's just do a quick
14:47 recap. left-hand side we're operating inside a project. This is a file in your
14:52 computer. So I was in my documents in a folder called agentic workflows in a
14:57 fresh folder called nadident builder. What we see here are we have different
15:00 files that we've created in this project. So we've got claw.md which is
15:04 kind of like the system prompt. It runs through what this project is actually
15:08 built for. We have connected our MCP server which is the mcp.json right here
15:13 which is the nin mcp and we've added those seven nen skills as you can see
15:17 here. expression syntax tools expert blah blah blah and those can be found in
15:22 here under skills which is what we connected from the GitHub repo and in
15:26 the middle what we're looking at is talking to Claude itself and as we're
15:30 talking to it basically makes to-do lists it makes plans and it tells us
15:34 what it's thinking so hopefully by now we're not too intimidated so what it's
15:38 telling us now is to restart cloud code in order to actually load those servers
15:42 and skills after restarted just test by asking can you search for the web hook
15:46 node using MCP So, let me just actually copy this right there. And then it says
15:51 the Naden skills folder was cloned for reference. So, they basically just
15:54 cloned it from the GitHub repo, which is why we see so much more stuff in here
15:57 than just the actual seven skills itself. And it tells us that we can
16:00 delete it if we don't need it since the skills are installed globally. So, right
16:03 now, I'm not going to worry about deleting that. But if we wanted to clean
16:06 some stuff up, we could obviously have Claude walk us through how to delete
16:10 just those things that we don't need. So, I'm going to go ahead and close out
16:13 of this conversation, exit out of VS Code, and as we open it back up, we are
16:17 already still in our folder. If we weren't in this one, we would just go to
16:20 file and we would open folder and hop back into it. But now, we're going to
16:24 open up Cloud once again. And I am just going to say, hey, just wanted to check
16:28 that you were able to successfully install and use the Nitn MCP and NIN
16:33 skills. And then I'm going to paste in that one test thing that it asked us to
16:37 do, which was, can you search for the web hook node using the NN MCP server?
16:42 So it says, I'll test the NN MCP connection by searching for the node.
16:44 You can see right here that it's actually using it. It's calling the
16:48 search nodes function. And it says, yep, I was able to find 13 results for web
16:53 hook. Also found Facebook ads, WhatsApp, Slack, blah blah blah. So it sounds like
16:57 we're all successfully ready to go with building end workflows. So, remember how
17:01 I said that if you are getting confused or if you don't understand how
17:04 something's working, you just ask. Well, here's an example of me just checking,
17:07 are you also able to create workflows in my instance, it verified with the MCP
17:10 server and it said, yep, here's information about your instance. Here's
17:13 what I can do. And then I said, do you also have the seven edit and skills that
17:16 we need to build workflows and it went ahead and it looked through everything
17:19 and it made sure we did have them. And it does in fact have them. So, now it
17:22 says ready to build some workflows whenever you are. So, I'm going to go
17:25 Building an n8n Workflow
17:26 ahead and do slash clear. We're going to clear the conversation and let's prompt
17:30 it with a workflow that we want to build. Okay, so remember earlier we
17:33 talked about the different modes. What I always want to do when I'm first
17:36 starting off with some sort of plan is obviously to go in plan mode so we can
17:40 make sure we have a really nice prompt before we go ahead and let Claude code
17:44 loose. So I'm in plan mode and I'm going to start to explain a workflow that I
17:48 want. Hey Claude, I'd love to build a workflow where we have an AI agent and I
17:53 want you to build this workflow in my end instance. I want this AI agent to go
17:58 off every morning at 7 a.m. and I want it to do research for me to help me find
18:04 out information for my business. I have an AI automation agency. So, I want to
18:07 make sure that every morning I'm getting insights as far as any AI developments
18:12 or um specifically AI automation related developments for um voice agents because
18:17 that's the niche of my agency. Once it has that information, after it's done
18:22 research, I want it to create an HTML styled newsletter for me and send it to
18:26 my email through Gmail. And I want two separate AI agents. I want the first one
18:30 that does research to use open router and to use chatbt 5. And I want the
18:35 second agent that writes the email and stylizes it in HTML to use open router,
18:40 but I want it to use anthropic sonnet 4.5. Okay, so that is our request. We are
18:46 going to shoot that off in plan mode and now it's going to come back ask us some
18:50 questions maybe and build out a plan and then we'll be able to go ahead and
18:53 approve that. And so this is exactly what we wanted. What it's doing is it is
18:57 creating a task. It's exploring through the MCP servers to search for different
19:01 nodes and then it's going to build us an implementation plan. So I'll check in
19:03 with you guys when that plan is all ready. All right. So it's a little bit
19:07 darker but it says I have a good understanding of the requirements. Let
19:10 me ask a few clarification questions. The first one is which email address
19:13 should the newsletter be sent to? I'm just going to go ahead and type in
19:18 [email protected] just to shoot that off. We've got the search tool. Which web search tool do 19:23 you want to use? Okay, we've got SER API, GINA, http. I'm going to go ahead
19:27 and say other. And I actually want to use perplexity. And then for time zone,
19:31 it says, "What time zone do you want to send from?" And I'm just going to go
19:34 ahead and say Eastern time, which I'm not even in Eastern time, but just to
19:37 make it easier, shoot off those questions. Okay, so we've got a pretty
19:41 comprehensive plan here. As you can see, it is our AI research newsletter
19:45 workflow plan. It's got some architecture, so the different nodes,
19:48 the schedule trigger, all this kind of stuff. The configuration details of each
19:51 of these. Obviously, we could read through all of this and make some
19:54 changes if we want. What I'm going to do is we're just going to go ahead and auto
19:57 accept this to see how well it did on its own, and we will see what happens
20:01 next. Oh, wait. Actually, I'm going to stop this real quick because what I
20:05 noticed is that it said what I'm creating is a single workflow file in
20:09 JSON. But actually what we want to do is have it created in our edit instance. So
20:12 I'm just going to go ahead and change this to bypass permissions and I'm going
20:17 to say that plan looks great. I want you to just go ahead and create that in my
20:21 edit instance and just give me a link so that I can go look at it. All right. So
20:24 there we go. You can see it had a to-do list with three items. One was to create
20:28 the workflow, one was to validate it, and then one was to report completion.
20:32 And so we have the link and then we have some next steps as far as connecting
20:35 your credentials, making it active, things like that. So let's click into
20:39 the link. Okay, we can see we've got the workflow right here with our schedule trigger,
20:43 our research agent, our newsletter writer, and our newsletter Gmail tool.
20:46 So, let me real quick just connect all these credentials. Okay, now that that's
20:49 set up, we're going to go ahead and run this. So, we're pretending that this
20:53 went off at 7:00 a.m. Eastern time. We can see that it's using its Perplexity
20:57 tool, its open router, and let me just also make sure this one was supposed to
21:01 be GBT 5, which this one's 4.1, so it messed up a little bit there. It filled
21:04 in tokens and temperature, which is kind of funny. This one was supposed to be
21:08 set 4.5 and it was sonnet 4. So it messed up a little bit on that actual
21:12 configuration, but that's not a huge deal. You can see that it's searching
21:15 perplexity. It's actually done three searches already. So if I click into the
21:19 research agent, we can see that we have a user message, which is to research the
21:23 latest developments. Actually, let me just make this full screen. Research the
21:26 latest developments in AI automation and voice agent technology from the past
21:30 week. Focus on these five things. Search for these four things and then compile
21:34 your findings. It also says to source citations, which is nice.
21:38 And then if we go into the actual system message, let's see what we've got in
21:42 here. You are an AI research analyst specializing in AI automation and voice
21:45 agent technology. Your role is to gather relevant information, be thorough, be
21:50 concise, prioritize actionable insights that would help an agency owner stay
21:53 ahead of industry trends. So, it's really good that it's making these
21:57 prompts pretty nice for us. Once this research finishes up, we will go over to
22:00 the newsletter writer. So, let's take a look at this prompt. Okay, cool. So, the
22:04 user message was already an expression because it knows that it needs all of
22:08 the research from the previous agent. And we can see that we've got it coming
22:11 through right here. And then it says, "Transform the above research findings
22:15 into a professional HTML newsletter. Follow the styling guide in your system
22:18 prompt." Okay, so we'll take a look at that styling guide in a sec. What we can
22:21 see here is the we get the actual results of the research from the
22:24 previous agent. So, we've got breaking news in AI. We've got voice agent and
22:28 conversational developments, AI automation tools and platforms, industry
22:33 trends affecting AI agencies, and new voice AI products, and then some
22:38 headlines and action for agency owners. Super cool. And then real quick, just to take a look
22:43 at the actual system message, it basically gives us a HTML styling guide
22:48 and some newsletter structure, which is great. So, it looks like that all
22:51 finished up and it got sent to the correct email, I hope, which was
22:54 [email protected] . It comes through as AI automation daily brief with the date and it's formatted 23:00 which is nice. And then we have the actual message and it even turned off
23:05 the append edit an attribution which I don't think I explicitly told it to do.
23:08 So that's pretty cool. But let me go to the email and let's take a look. Okay.
23:12 Wow. Look at this. We've got our AI automation daily brief for January 14th.
23:17 It's obviously stylized in HTML. We've got an executive summary with the Aentic
23:21 AI market exploding. We've got top stories. So, we've got some headlines
23:24 like the agentic AI market is set for growth, enterprise agent ops, hardware
23:29 breakthrough. We've got some voice agent focus because that's what I told it was
23:32 my niche. So, we've got emotionally intelligent agents, global support, new
23:37 voice AI integrations, some quick hits, some action items for agency owners. And
23:41 then at the bottom, we've got a bunch of sources, which is pretty cool as well.
23:44 So, obviously, this was our oneshot prompt for this workflow. If we wanted
23:45 Using Claude Code Deeper
23:48 to iterate upon this, we could go ahead and come back into VS Code with our
23:51 cloud code and say, "Hey, this workflow is great. Can you make these changes to
23:54 the prompt or could you switch out some things here?" And it would actually be
23:58 able to take that instance, change it, and then we'd see those changes
24:01 reflected in our end cloud. This workflow obviously wouldn't take me, you
24:05 know, more any more than an hour if I was building it myself in Naden, but
24:08 hopefully you can see if you're a beginner and you need to help brainstorm
24:11 through some things, cloud code's going to help you out a ton. And also, if you
24:15 think about as you scale up to some more complex use cases where maybe you need
24:17 to think about how you're going to handle the data and things like that,
24:21 using cloud code and plan mode is really, really going to help you out
24:23 because it's going to ask you a lot of questions that you may not have thought
24:27 of yourself when you're building out your workflow. And something else that I
24:30 talk a lot about in my plus community is the importance of wireframing things
24:33 out, as you guys could tell earlier when I did this. And so I imagine if you
24:37 wireframed out a workflow and you gave that screenshot to a cloud project like
24:41 this, a cloud code project like this, it would be able to do a lot better too
24:44 because it already knows kind of like the general architecture and structure
24:47 that you're thinking of for your NN workflow. Like I said, that's something
24:50 I talk a lot about in my plus community with full courses on NN and building
24:54 agencies and stuff like that. So if you're interested, the link for that
24:56 will be down in the description. So let's just do a quick recap of what we
25:00 actually talked about today. We got familiar with VS code and cloud code and
25:04 we understand that the lefth hand side is our project where we have different
25:07 files and we have basically a system prompt that explains what this project
25:11 is going to do. We've got the middle section where we actually talk to either
25:14 we look at our files or we talk to cloud code itself. We have different modes to
25:19 talk to cloud code like bypass permissions, ask before edits, edit
25:22 automatically or plan mode which is really important when you're actually
25:25 trying to figure out what you want to do. And then we talked about how you can
25:29 give it access to MCP servers, skills, and rather than trying to figure out how
25:32 to configure it yourself, just make cloud code do the research. And
25:35 something funny about cloud code is sometimes it'll do the research and say,
25:38 "Hey, you know, I can do this for you." Or you could just install it yourself.
25:41 And a lot of times you just have to say, "No, you do it." Like, "I don't want to
25:44 do anything. I just want you to do everything for me." And it will do it
25:48 all for you. So key lesson, don't be intimidated by cloud code. I'm going to
25:51 have a lot more content around this coming. So, please let me know what type
25:54 of use cases you guys want to see when it comes to automating things with cloud
25:57 code. The thing you need to remember is all that matters is can you communicate
26:01 clearly what you want as if you were talking to a human developer or a human
26:05 project manager or even just a regular human. If you can communicate it clearly
26:08 and if you can ask questions when you're confused, you will be fine with cloud
26:13 code. And once again, the interface is going to get better, the models are
26:16 going to get better. So the earlier you jump on this, the easier it's going to
26:19 be when inevitably every person in the world starts using cloud code. Okay, one
26:23 final thing that I wanted to just make sure you guys were aware of is it has
26:28 its limits. So obviously I was trying to get to that line and see how far I could
26:31 push it and I was able to build some really complicated workflows and some
26:34 cool agents, but I tried building one where it was going to be like 60 nodes.
26:39 And what ended up happening is it basically like throttled the API and it
26:43 just like became unresponsive. And this was like over an hour ago. And my N&N
26:48 cloud is still down. And so the thing is here, NAN's cloud server is what's down
26:54 specifically for my domain here. But if I open up a different cloud instance
26:58 that I own from NN, that one's working. So there's an error going on here. You
27:01 can see it's 522. And I'm basically, I think, going to have to reach out to
27:05 them and have them like re reboot the server or something like that. If you
27:07 were doing this on self-hosted, you'd be able to reboot it yourself and you'd be
27:10 good to go. But this is something I wanted to at least let you guys be aware
27:15 of is that if you do make some massive crazy requests and it's timing out and
27:18 then it's trying again and it's just like, you know, just trying to do way
27:24 too much. This may happen. So, um I just felt like I was obligated to at least
27:27 tell you guys that. So, anyways, and if it helps you guys out, I will drop this
27:28 FREE Resources
27:30 claw.md file, which keep in mind I oneshotted with plan mode, but I will
27:35 drop this in my free school community if you want to grab that for free. The link
27:38 for that is down in the description. It will look like this. All you have to do
27:41 is get in here and I'll have a post associated with this video and like
27:44 let's pretend it's this post. You would just see right here I would attach a
27:49 JSON file or a doc with that exact cloud.md file in there so that you guys
27:52 could just paste that over to your cloud code if you want to play around with
27:55 this use case. But anyways, that is going to do it for today. So if you
27:58 enjoyed the video or if you learned something new, please give it a like. It
28:01 definitely helps me out a ton. And as always, I appreciate you guys making it
28:04 to the end of the video. So I will see you on the next one. Thanks so much,