How I Fully Automated My LinkedIn Posts! (Claude Code)

本站添加: 2026-05-09

[00:00:00]Today I'm going to show you exactly how I was able to fully automate all of my LinkedIn posts just by using Claude code. So on the screen now is a multi-image LinkedIn posts that I 100% generated using my Claude code. And so I've got four fully generated images right here as I scroll through them that were 100% generated by Claude code. I really didn't have to give it too much work on my end for it to be able to generate all of these images. All of the text in all of these slides is all perfect as well. There's no AI generated sloppy text in here. And at the end, we've got this call to action as well, which says, "Comment CRM, and I'll DM it to you." And so, this will obviously help out with engagement, getting people to comment. And once again, this has all been done automatically. I just gave it a brief idea as to what I needed a LinkedIn post for, and it was able to generate everything. It came up with the call to action to create for this post as well. And alongside all of these carousel images added into LinkedIn post, you'll also see that we have the actual LinkedIn post content as well.

[00:00:55]But honestly, from my experience, using LinkedIn and making LinkedIn posts as well, really getting these image carousels properly done is one of the best ways to boost engagement, especially with the call to action at the end here. Here is another example of a LinkedIn post once again that I fully generated. You can see here we got the main page. We then have multiple different slides to this as well for different elements a part of this post.

[00:01:17]This particular post has six total slides in it. And once again, at the end, we have comment skills for my full cord code setup once again for that call to action to boost the CTA. And once again, we got the full-on post text generated alongside this as well, which is all relevant with the actual images that are generated. And then to actually make the post, Claude Code connected it into this tool called Buffer, which is just a LinkedIn post scheduling tool.

[00:01:41]And you can see here, this is one of the posts that I just showed you, which is scheduled to send out tomorrow. You can see that it has added all of the different slides here as well. Added into this LinkedIn post alongside all of the copy with it as well. And so I obviously didn't have to manually add any of this in. It just added this into the scheduling system. And now without me having to do anything, this is all ready to go and we could obviously push it a bit harder to then start to make a post every single day for us. And it will be all fully automated. And so before I dive straight into Claude code and showcase exactly how I built this system, I just wanted to let you know that I've also fully open sourced this entire project into GitHub. I'm going to be adding the link to this GitHub repo with all of the code into my free school community, which you can find and linked at the top of description. Essentially, you'll just need to take this link right here, send it into Claude Code, and just tell it to build the system for you. So that will be the easiest way to get the system up and running. But I do recommend watching this video still to understand the fundamentals and to and to understand obviously what you're actually building here and how it works in the background. In just a second, I'm going to go through my entire Claude code setup right here and just showcase everything about it and how it's currently set up. But before I do that, I'll just give you a little bit more context as to exactly how this system works so that when we start to build it, you have an idea as to what I'm actually doing. So this is just on the GitHub page. I've created a readme file with a mermaid diagram right here kind of just walking through exactly how this system works. So once again, if you want to get access to this GitHub file, I'll have it in my free school community. So if you want to have a read through this document as well, you can do that as well. But ultimately, I've just called this the Claude LinkedIn auto poster. So it's an end toend pipeline that will turn any kind of topic and a few brand reference images into a multi-slide on brand carousel and schedules that to LinkedIn as a native document post as well. This is built using OpenAI's brand new GPT image 2 model. We're using Cloudinary to store all of the media and all of the posts. And then we're just using a couple of other random libraries which I'll talk about in just a second.

[00:03:33]But ultimately how this first works is that we just have to provide it with a generalized topic. So for me, I provided it with some YouTube video topics that I'd already created before. So I pretty much just provided it with the title of a YouTube video that I had created previously and it was able to use that and reference that to be able to create the slides. But you could pretty much just give it any kind of topic and it should be able to figure out what you want out of it. Alongside this, I uploaded some style reference images.

[00:03:56]So, I've got some images of thumbnails that I've created before, but if you have any style reference images, you will be able to just upload them to this system. And that's how it's going to be able to generate consistent and onbrand carousel images. You can see here with this generation, every single one of these images are very similar in how they've been generated. It's using all the same colors. The text font is exactly the same. It's using the sort of yellow box in the background. It all looks very consistent in its design and same goes for the second post that made as well. Very consistent in its design and they all look very similar. And that is obviously by design with the reference images that I've uploaded. So this is built into the system and 100% would recommend you doing this as well.

[00:04:32]So all of that information then gets sent off to OpenAI's GPT image 2 model.

[00:04:36]If you're not familiar with OpenAI's image 2.0 model, no doubt about it, this is the best image model that is out there when it comes to generating pretty much anything. So, I would recommend uh no matter what you're doing, definitely choose this model. It's the best one out there. I did recently create a video around comparing this to the Nano Banana Pro model and this beat it pretty easily. So, this is then going to get initiated. We're going to be generating multiple of those images for those carousol slides. Then, we're going to go down two different pathways. One is going to be adding the PGs locally onto our computer. Once again, we're running this on clawed code, so the files are going to get stored locally on our computer. Then once it's created the multiple of those images, we have a setup that is going to stitch all of those images together. Usually OpenAI's GPT image 2 model will just output PGs or JPEG. But when uploading those carousol images on LinkedIn, it essentially has to be a PDF file in order for it to do that. You can't just upload multiple different PNG files. So this system will take all of those PGs and combine it into a PDF. And once that's created, and it's got the carousol PDF file, we're then going to be uploading this into a tool called Cloudinary. So this right here is cloudinary. I've been using this to store all the images, videos, and files that claude code has been generating.

[00:05:43]And the main reason that we do need to do this is that later on in the workflow here, we are going to be sending this off to Buffer. Buffer is just a social media post scheduling tool. We're not actually able to do what we want to do natively through the LinkedIn APIs very easily. And so using a tool like Buffer just makes it a little bit easier. But in order to actually do this schedule post through the buffer APIs, we require a public URL of that PDF. And so by storing the PDF file here in Cloudinary, this is essentially just going to give us a URL that we can publicly access so that when we make that API call to Buffer, it's going to be able to retrieve this PDF file very easily as we can't just host this locally on our computer. And the same thing goes for all of the individual PNG slide images.

[00:06:25]If I come back to my Cloudinary account, you can see all of the images that it is manually uploaded into here as well. So, one of the really big benefits to using Cloudinary to actually store all of this content as well is that they are automatically going to be tagging all of these images using AI. So, if you are like me and you're creating a lot of content nowadays using AI, whether that's YouTube thumbnails or LinkedIn posts or just anything, you probably know how hard it is to manage all of that content. In my downloads folder specifically, I've just got like hundreds of images that have generated and it's pretty much all just scattered all around the place and it's all just getting a bit messy. But if we do go ahead and use a tool like this, it just makes it a bit easier to actually find all of these assets, tag all of these assets. And if I ever need one of these posts or ever need one of these images, I can literally just type up here to find it. So let's say I wanted the original image of the Claude mascot, all I would have to type is just mascot in the search bar right here. And now it has found this original image of the mascot that I had previously uploaded in one of my other queries. And so the same thing goes for literally every image right here. Just allows us to very easily find any assets that we generate.

[00:07:23]And this has really good APIs. So I've pretty much fully integrated this into my claude code. So whenever I generate an image of any kind, it just throws it into cloud and area here. So if I ever need to find it again, it's very easy to do so. And also, if you are a content team, you've got other people working for you, you will obviously need some sort of a cloud hosted area to put all of those assets into as well. So, everybody can interact with those files.

[00:07:43]So, for example, if I wanted to share some files off to my thumbnail designer, he is now very easily able to log into his account and see all of these assets.

[00:07:51]And if he wants to use them or modify them in any way, it's very easy to do that. They've also got a bunch of other AI features. This one right here that they're showcasing pretty much detects certain elements in the images as well.

[00:08:01]So, if you ever generate something that has, you know, an iPhone in it or a cup in it and you want to later retrieve that image or you're thinking of that image and you can't find it, it's going to be able to automatically detect all of that stuff and all of these assets you're going to be able to retrieve and find things very easily. Anyways, that's what we're using Cloudinary for. Right after this, it's then going to get sent off to the Buffer API to then make that LinkedIn scheduled post. And so, this once again is Buffer. I'm currently using the free plan. I think you can do quite a few posts if you are just going to be connecting into LinkedIn on their free plan. You don't even have to upgrade. And pretty much all I did was I told Claude Code that I've signed up for Buffer and I want it to do the scheduled post through this tool. So realistically, you don't even have to use this interface at all. Claude Code is going to manage everything for you and you don't have to worry about any of this. All right, I'm now going to scroll through my Claude Code instance right here and just showcase all of the messages that I sent off to my Claude Code to build out their system so you can get a super transparent idea as to everything I did in order to get this set up. So originally in this folder here, I wanted to get Cloudinary set up so that I could start to store all of these files in a publicly hosted environment. So that's why my first prompt here just said, can you set up Cloudinary media storage here? So when generating images and videos, they also get stored in Cloudinary and I can surface the media easier later. Once I set that off, it did some research about Cloudinary, came back and then just asked me for my Cloudinary account details. And then it just quered me as to how the media is going to reach this as well. And I pretty much just told it that any assets that get generated in this conversation or in this folder, I want to have automatically uploaded into Cloudinary to be retrieved later. Once I did that, it had access to everything that it needed. It found all of the APIs online and it was able to get the entire setup running. It then just needed my cloud name from Cloudinary, which I just went into my account and found. Once I done that, it was all pretty much set up and ready to go. I tried to send it an image to get that uploaded and just test if the connection was working. But just a note from my end, you aren't actually able to upload an image into the claude code instance right here and then have it uploaded. The best way to do this is to actually put the image itself into your folder so I can actually see that image. And so once I did that, it was working fine. It was able to find that image and then it gave me back the exact URL as to where that image was stored.

[00:10:03]So once that was all set up, I then started the process of building out the Instagram carousle post creator. I said, I also now need you to become an Instagram carousel post creator. Now, obviously, this was for LinkedIn, but I said Instagram originally and then changed my mind that I thought it was actually better suited for LinkedIn. But anyways, it's kind of the same thing.

[00:10:19]Anyways, I then said I've uploaded some reference images from my YouTube videos for you to take style guides from, but the style should still be a proper Instagram carousel post for the topic. I then told it to please use the Open AI image to model for your image and store these images in Cloudinary. And then I said an Instagram carousel needs to be three to four images that all flow to each other for that post. So, the reason I said this is that we don't just want these to be all individual images that don't relate to each other. I think that this just helps to make these posts a little bit more engaging. If I take a look at these images right here, you can see we got the swipe at the bottom right. It just came up with doing this, but that was obviously based on the prompt that we gave it here. And then you can see in the top left, we've got the numbers for the one of five. So, obviously, it has counted each of these slides and it's done that. So, it kind of engages you to then look at the third one and then the fourth one and to keep going through this system. Now, as for the reference images, I said that I uploaded some reference images. You can see here we have this style guide folder right here. And this is just where I uploaded all of my YouTube thumbnails.

[00:11:15]So, this is inside that folder. You can see here all the images that I uploaded for this. Once again, I already had these YouTube thumbnails, all AI generated, but I've just been using a different system to make them. So, if you do have any reference images to use and take examples from, I would recommend uploading as many of these as possible. So, that way we can get consistent designs that are going to follow along with the style guides that we have. Once this was done, it pretty much said that it is all good to go. It had a couple of questions for me.

[00:11:40]Obviously, this is always going to happen. Just follow through and and answer these questions that it gives you. It first just asked me what the topic was about, so I just gave it one of my YouTube video titles for it to create the topic around. And then the second one was a slight issue. It didn't seem to find that OpenAI image 2 was a real model. just because the OpenAI image 2 model is so brand new. Not a lot of these AI models have yet been trained on the fact that these models have been released already. So I just told her that GPT image 2 is now out and to do some better research on that. And then lastly, number three, it just asks for my open AI key. If you are going to be giving Claude code any of your API keys to practically anything, I usually wouldn't go ahead and just paste it in the chat for it to get access to. The reason for this is just security. It's not a safe idea to do that at all. that can get leaked very easily. And if it's in the code that can be found very easily. So I would just recommend asking it to give you some instructions if you're not yet sure on how to add an API key, just ask it for instructions on where to put that API key. And that's going to be much more secure. So four slides were now completed for that particular title that I gave it. And you can see here it provided kind of a breakdown as to each of the slides that it created as well as a view button for me to find that. And because we already set up Cloudinary before, it's practically already done for us.

[00:12:46]Cloudonary is now storing all of those images and everything is now locally stored as well if we need it. And so at this point we're practically done when it comes to the generation part of this.

[00:12:55]But obviously the last part of it is just getting that LinkedIn post text generated as well. I went ahead and just said thanks. Can you write the LinkedIn post content for this as well? And so the first post that generated wasn't too bad. I would say there's a lot of improvements that can be made obviously with writing content. But one thing that I did do later on for the next post is that I actually found a skill online specifically for LinkedIn post generation. And so after this, I got it to create a new carousel, which was the second one that you saw around this particular video. I actually gave it the link to the video as well. And once this had generated and I asked it to create the post, the post I got back was pretty terrible to be honest. It was just using these emojis and it kind of looked like the most AI generated looking text post you'd ever seen. And so I wasn't very happy with it. So what I did is I added in a new LinkedIn post writer skill file and taught it to install it here. And so this is the LinkedIn post writer that I found on MCP market. And so all I did is I went ahead and clicked download skill, downloaded it for claw code, and then uploaded that directly into my claude code instance. It then looked at the skill file, downloaded it, and then was able to use that for the next generation that it did. If you do want the link to this skill on MCP market, I'll put the link in my free school community alongside the other stuff as well. But ultimately, here's the new post that we got back from using that skill. I do think it is quite a bit better in terms of the structuring and you can see some of the reasoning that it had as well.

[00:14:08]But overall, I still didn't put a whole lot of effort into the actual generation of the styling here of the post. I would say that a lot of LinkedIn content is quite sloppy anyway. So, it is something to be about. So, it is something to just learn over time. So, once that had been generated after this, I now wanted to get published and scheduled off to my LinkedIn account. So, pretty much I just told it, "Thanks. Are you able to schedule these posts directly to my LinkedIn account?" Once I sent that through, it gave me a few different options for how I could potentially do this. The first one was LinkedIn's nativeuler. Although, this was just for me to do it manually. It was just telling me that I could go ahead and manually do it myself, which is true if you're doing a low volume amount of posts, but we want to fully automate this, so that's not really going to be an option. The second option was an NN workflow that I could potentially use, but that was going to be a little bit of a hassle to set up. And then the third option was some dedicated tools, which said that it was best for native carousels at scale, which is exactly what I want. And then it gave me three different options with Typefully, Buffer, and Publer. So, I had never heard of these tools before, although I just asked it which one is the best. And it kind of gave me a breakdown of the different platforms here. And so, I just said, "Thanks for that. I'm going to set up a Buffer account right now, and can you connect into their API to schedule a LinkedIn post for tomorrow?" Once I did that, it looked on the internet, found all the APIs related to Buffer, set up all of those APIs as well with all the connections, and then I just answered some of the questions as to how I wanted these posts made. As this process was going, it did ask me to make a quick change on Cloudinary to allow the delivery of PDF and zip files because we now are going to be combining all those PGs into that PDF file. We did have to enable this. Once that was done, it was all good to go. It was able to download the PDF. It was able to send that through the API to Buffer and the post was able to get scheduled. And so this is exactly then what popped up on my Buffer account. I told it to schedule the post for tomorrow morning and that's exactly what it did. It's added in the entire PDF file. You can see PDF right here. And if I view this, you can actually see the PDF file of this as well. You can see all the text related to the post as well. And at this point, there's literally nothing else I need to do. I can come over to this claude code instance. I can give it a brand new video ID or a brand new post ID. It's going to be able to create all the images. It's going to be able to create all of the text content as well. And then it's going to be able to send it off and schedule it on LinkedIn in literally one message that I send it.

[00:16:13]And another area to note is that if you ever need to make changes to any of these carousel slides, if anything doesn't work the way you'd expect it to, one thing that you can do is use some of Cloudinary's AI tool to quickly modify or make changes to any of these post images. Essentially, what we can do is just come here, modify a file, and then just tell Claude Code that we've modified one of the files. It can then automatically retrieve all these files, turn it into the PDF file once again, and then send it over to Buffer as that scheduled post. So if you need to make any modifications like this in changing colors, adding effects, doing a lot of AI transformations as well with generative fill, generative removal, background removal, all this stuff for making quick changes. I hope this video was helpful in fully automating your LinkedIn posts. Once again, if you want to get access to this entire project to be able to take and literally just get Claude code to clone for you. I'm going to have this entire GitHub repository of the entire project in my free school community for a completely free download. And just for my last note, I've recently discovered that 90% of you watching right now are actually not subscribed to the channel at all. So if you could make sure that you are subscribed to the channel, just scroll down a bit right now and check if you are subscribed. Subscribe right now.

[00:17:14]I'll keep pointing down so you can click the button and you have enough time to do it. I've got a lot more Claude Code content just like this planned up for the next few days and I don't want you to miss it. And so if you do subscribe, you will get notified and you'll be the first one to find those