Meet the GitHub Copilot app: Your new AI desktop assistant

[00:00:00]Working with AI agents to help generate code can feel pretty linear, often with one chat per project at a time in an interface that can feel isolated from the rest of your work. The GitHub Copilot app is built differently, centered around Copilot and GitHub, allowing you to easily manage your outstanding tasks, your current tasks, and assign new tasks to a team of AI agents as inspiration strikes. Let's explore what makes GitHub Copilot app special.

[00:00:33]The Copilot app is the agent native home for your work on GitHub right on your desktop. If you haven't already worked with AI agents before, agents can go beyond just generating code. They can run builds and tests, create PRs, and even access external services. You can control their level of autonomy depending on your needs and scenarios.

[00:00:55]From the home screen, I can see the core chat window where I can start new agent sessions. I can select the project that I want, the mode that I want, and the model that I want. Now, the list of projects here is only going to show me the projects that I've worked with previously. If there's a new project that I want, I can select add project where I can choose to pull one locally from GitHub or based on a URL that I give it. Now, I want to work with Tailspin Toys and I want to add in a breadcrumb to our board game crowdfunding app that we're going to use throughout this course. Let's add the prompt, "Add in a breadcrumb to show the most recent five games the user has visited." When I send that, Copilot is going to get right to work. Let's go back to the home screen, though.

[00:01:48]If I go back to the home screen and scroll down a little bit more, what I can see is the list of unmerged pull requests and open issues that I have, which allows me to quickly start new sessions based on work items. Down even further, there are some tips on how to utilize the tool and a link to the documentation. Speaking of the documentation, the biggest tip I can provide to quickly get up to speed, besides of course watching these videos, is to ask Copilot about Copilot. If I create a brand new quick chat, I can ask how to work trees work with Copilot app.

[00:02:27]And Copilot will then give me the response. Now, I'm not limited to just asking about documentation inside of here. That if I want to ask about a project, I could do that as well. For example, I start up a brand new quick chat and then I can select that Tailspin Toys and then ask how does this project work. You'll notice that I'm able to spin off all of these different sessions without having to wait for any of the others to complete. I can then of course go back to the ones that I had previously, like this one right here, which shows me how work trees are going to work inside of Copilot app. Now, speaking of these different sessions, let's go to the one that I just launched a few moments ago where I asked Copilot to add in that breadcrumb. If I scroll up towards the top, what I can see is that Copilot started by exploring the project and determining what it is that needed to be done. It opens up different files and then eventually here will create the code and run our tests. Now, I requested a UI update and as a result, I want to actually see the UI to confirm that things look good. Rather than opening a new terminal to start the site and launching a new browser, I can create a browser canvas right here inside of the Copilot app. Now, a canvas is a shared interactive service built around actual work artifacts like a plan, a Kanban board, or a checklist.

[00:03:59]Or, in my case, it's going to be this app. That way, I don't have to just trust a wall of text. I can work iteratively, steering Copilot based on something I can see. So, I'm going to utilize the create canvas command here, and I'm going to say open this app in a browser canvas.

[00:04:21]Copilot will then get to work by starting the app and then eventually opening up my canvas here. With my canvas open, I can now interact with my site and see how everything looks. So, let's open up the Binary Frontier here.

[00:04:35]Let's go back home, and then now I can see that it's properly displayed here.

[00:04:40]But, maybe I want to change the font color that I've got here. I can select pick and polish.

[00:04:46]I can point at Binary Frontier, and you'll notice that actually brings that into context with my chat, and now I can ask chat to make an update. Let's make this text a bit lighter.

[00:04:59]Copilot will then be able to locate that item on the page and make that update.

[00:05:04]That allows me to work back and forth and work together with Copilot to make sure that I'm getting the output that I need. Everything looks good, so it's time to create a pull request. That PR, of course, still needs to go through the standard CI or continuous integration flow. And as any developer will tell you, just because everything runs locally doesn't mean it'll work when it's pushed. Just as Agent can resolve local test failure, Agent Merge can monitor everything on my repo as well.

[00:05:33]It can address any feedback from the rest of the team, fix any CI issues, and resolve merge conflicts. Basically, the Agent can make sure everything is ready to merge once I hit go. I can start an Agent merge by selecting the drop down here and then choosing agent merge and then hitting the button.

[00:05:54]Co-pilot will then go ahead and look through my project and any special instructions that I might have. In my case, I have a special skill which tells Co-pilot to create an issue. We'll talk about skills in a later module. That it's going to create that issue first.

[00:06:11]Then it creates the PR for me. With our pull request created, the agent will then monitor how it's flowing through that continuous integration flow. And you'll notice when I hit the drop down that I can allow the agent to address any reviews, to fix any CI failures, and to resolve conflicts. And if I decide that we're ready for prime time, I can even allow the agent to merge that pull request. So, it will wait until our CI process completes and then merge that code into our repository and into our main branch in this case. Now, that's probably the most common path through Co-pilot app. You can use quick chats to get quick answers including about Co-pilot app itself. You can create sessions on your repos with each session using work trees to be independent of one another. And you can even use agents to help you manage your PR and CI flow.

[00:07:07]There's of course lots more to explore which we'll see in the rest of this video series. We're going to dig into the functionality a bit deeper, explore, and introduce parts we haven't yet touched on. As always, the best way to learn is to get in and play with the tool yourself. You can download it at gh.io /app. Try it on a repo you know, maybe with a task that's been sitting on your backlog for a while now. You might be surprised how much it gets done while you're still learning the tool.