Transcription
What if I told you you could have an AI coding agent working from your phone, your computer, or wherever you are, even while you sleep? That's what the new GitHub actions integration does with Cloud Code. Today, I'm going to show you how to set it up and put it to work. Let's go.
What is GitHub actions? If you're using GitHub to manage any code that you're working on with Cloud Code, you can tag @claude in GitHub and it will work in the background to solve whatever you throw at it. This is a video that Anthropic released about a month ago that shows what this GitHub integrations looks like in action. They're just tagging @claude and asking it to solve this issue in the background. You press enter in GitHub and it just starts working. You don't have to do anything. Claude code comes back with a pull request that you can review and then merge into your code.
I'm assuming most of you know how to use GitHub. But for any newcomers to the coding world, if you have no idea what a pull request is or how to use GitHub, don't worry. I've got you covered, too. It's super easy to set all this up. So, let's get into it.
Open up the code you're working on. Whatever code editor you're using. You can see here I'm using Cursor. Whatever repo you're in needs to be connected to GitHub. So if you haven't done that and don't know how, I'm assuming most of you do. So if that's the case, skip ahead 20 seconds. So I'm not preaching to the choir here. But if you're a total beginner and you're starting from scratch, log into GitHub and create an account. Takes less than a minute, super easy, and create a repository. Once you do that, clone the repo locally or use the GitHub desktop app to clone it locally. These are all simple workflows with simple instructions on GitHub how to do this. It is foolproof, trust me.
So once we're in the code editor, launch Claude by typing `claude` and type `/install GitHub app`. So it'll ask you which repository we want to use. I'm going to use this one. Press enter. Now it will redirect me to my GitHub account here. And it will ask you to install and configure this on your GitHub repository. Okay. Now I've already installed the GitHub actions app in my GitHub repo. So it will take you back to your code editor and press enter once you've installed it and then it will tell you which workflows it's installing.
So this first workflow right here, `Claude code`, that's basically what I was showing you earlier. You can tag @claude in issues and PR comments and it will work in the background to solve the issue or PR comment for you. And then the second workflow called `Claude code review`. It's basically just a way to automatically review every new PR or pull request that you're running through GitHub. So, both of these features we're going to install. Press enter.
For the next step, this is the biggest new unlock in my opinion. If you recall from my first video on Cloud Code where I briefly mentioned GitHub actions, it's been out for a while, but you had to use an API key at first to use it, which in my opinion is a huge waste of money if you're paying for Cloud Code Max at $200 a month until now where Anthropic lets you use your Cloud Code subscription. And that's what this long-live token is talking about here. You can just use your subscription. You won't be charged on a per-use basis. It's part of the subscription. And you don't need a max plan either. You can be on the pro plan to use this. So start using it today if you're a pro subscriber and up.
Then you'll be redirected to your Anthropic account to authorize the long-live token. Click authorize. You can close the window. Once you close the window, it will actually create a pull request. And this is the first thing that will happen when you install GitHub actions in your repo. See, it's created a pull request called `add cloud code GitHub workflow`. So just click create pull request and it's going to run some checks. And once it passes, we can merge this pull request into our code, which will now enable us to use Claude code GitHub actions in the background. So I'll come back when this is finished and we can start with our first workflow.
Okay, so you can see the second cloud workflow has executed on this pull request which again just adds the GitHub actions to our repo and it's it's reviewed all the code that is generated to add this integration. So once that's finished, you can just say merge pull request and now we can start using the GitHub actions.
So, here's an example from a codebase I'm working on. This is the current mockup of the user interface that I've iterated with Vzero, and I've downloaded this front-end code into the codebase that you just saw on GitHub. The issue that I want to address or the feature that I want to add here is the waitlist feature. These are just placeholders right now. These buttons and forms don't work. This button up here doesn't work. So I'm going to create a new issue in GitHub to implement these features and to add in some back-end features that are necessary to drive this and have cloud code do all the work.
So I have created an issue description here in GitHub to add the join waitlist button functionality and to add SendGrid and database functionality so that this can all work. So then I'm just going to go down here and comment on the issue and say `@claude please implement features outlined in this issue`. And now it will start working.
Okay. So you see I actually had to send another request. It said to implement this feature please because I was having some trouble actually getting it started. But now it works just fine. And you can see Claude is coming up with a plan in the background to implement the features outlined in the issue.
So, while this is working in the background, let's check back in to see what Claude Code comes back with and see if it implements the features correctly. But the coolest part is that you can actually do this on the GitHub app through your phone on the go. So, you can see I have the GitHub app on my iPhone and I can look through this issue and interact with Claude Code here. I can see the progress that Cloud Code is making on the go and I can even interact with it and create more issues and more comments just by tagging @claude the same way we do on the computer.
Okay, so it took it about 10 minutes, you can see, but it finally finished. All right, so the next step you do after Claude Code finishes this is to go to your branches. You can see Claude creates a separate branch to work on your code. So it's not making edits in the main branch. You create a new pull request and Claude will review the code to check to see if it's okay. And then once that's done, you can merge the pull request.
So now we're going to see if it implemented everything correctly. So, I've run the development server, open this up in my browser. So, let's see if I can successfully add my name to the waitlist. Okay. And if I go to my Neon console, I can see that I've successfully been added to the waitlist. Now, there's ways to probably improve the UI here by giving a confirmation in the browser and not just sending a confirmation email to this email address, but Cloud Code basically oneshotted this through the GitHub actions integration, which is totally cool. I didn't have to interact with this at all. This is awesome. And if I open up my email client here, I can see I successfully got the waitlist confirmation email as well. So Neon, the database is working to store all my emails in the waitlist database and the SendGrid API is working to send confirmation emails to everyone that signs up for the waitlist.
So again, this is great. Cloud code GitHub actions essentially gives you a second developer who works for you around the clock wherever you go and wherever you are. For anyone who wants to move even faster with AI coding agents, try this now. If you're already using Claude Pro or Max, there's no reason not to set this up today. It just takes a few minutes and you'll immediately start seeing some benefits. Let me know in the comments what workflows you're working on with the new GitHub actions features. I love seeing what you guys are working on. And if this video was helpful to you at all, follow this channel for more AI tutorials. And if that's not enough for you, you can always follow me on X. See the show notes below for more real-time updates. Thanks for watching and I'll see you again next.