Exploring the agent-first GitHub Copilot desktop app | GitHub Checkout
Chapters8
This chapter explains what the GitHub Copilot app is and how it moves from CLI-based workflows to a user-friendly UI that lets you manage code, issues, and projects with quick chats, PRs, and automated merges. It highlights how users can start sessions from issues, view and run changes, and let the agent handle PR creation and merging after CI passes.
GitHub’s Copilot desktop app blends agents, automation, and chat to manage code, PRs, and project work in one UI.
Summary
James Clancy showcases the GitHub Copilot app on GitHub Checkout, revealing how agents help manage code, issues, and project workflows from a single UI. He explains starting a quick-chat session from an issue to have an agent fix code and push a PR, with automatic PR merging once CI passes. The app supports agent-based automation in the cloud, including triaging issues, syncing feedback across repos, and even transcribing meeting notes for customer docs. Working trees enable multiple agents to work on separate copies of code tied to different sessions, making prototyping and testing more streamlined. James also highlights that existing CLI capabilities map directly to the app, with future plans to support local models and user-supplied models. You can open work in your editor of choice (VS Code, Xcode, Visual Studio) or stay entirely within the Copilot app for development, testing, and UX feedback. The interviewer and James emphasize real-world use cases across product management and software development, plus upcoming features and community feedback channels. The overall message is that this is a rapidly evolving, highly integrated tool designed to reduce manual setup and context switching while extending agent-powered automation into everyday workflows.
Key Takeaways
- Starting from an issue, you can create a session in the Copilot app, have an agent fix code, and view the changes before choosing to create a PR.
- The app can automatically monitor a PR after it’s created and perform agent-based merges once CI and reviews are ready (agent merge).
- Automations live in the cloud and can run triage, email checks, and meeting-note transcription on a daily schedule, all without manual cron setup.
- Working trees let multiple agents work on isolated copies of the repo, enabling simultaneous prototypes on different folders tied to their sessions.
- The Copilot app exposes the same models as the CLI, and roadmap includes support for local or frontier models to reduce token usage.
- You can open code in your preferred IDE (VS Code, Xcode, Visual Studio) or stay inside the Copilot app for end-to-end development and testing.
- The app scales from developer tasks to broader PM management by automating feedback triage, repo refreshing, and cross-repo coordination.
Who Is This For?
Software developers, frontend/backend engineers, and engineering managers who want a unified, agent-powered workflow for coding, PRs, and project management without constant context switching.
Notable Quotes
"The GitHub app just exposes quick chats as well as repositories for your source code or for even project management type things."
—Intro to what the Copilot app exposes and its scope beyond coding.
"So we’re able to, yeah, just take an issue, click on it, start a session. And from there, it will actually go through and do the work and fix the code."
—Demonstrates how to start a session from an issue and have the agent fix code.
"No more sitting there baby sitting CI, which is really nice."
—Highlights automated PR monitoring and merging after CI.
"Working trees… will create a copy of the source code that only that agent's going to work on."
—Explains the working tree concept enabling multi-agent parallel work.
"If you can do this in the CLI, you can do it here. The models are the same."
—Affirms CLI parity and model availability in the app.
Questions This Video Answers
- How does the GitHub Copilot app handle automatic PR merging after CI passes?
- Can I use local or frontier models with the GitHub Copilot app in the future?
- What are working trees and how do they support multi-agent development in the Copilot app?
- Can I open Copilot-created code directly in VS Code, Xcode, or Visual Studio from the app?
- What are practical automation use cases for the Copilot app beyond coding (e.g., triaging issues, managing waitlists)?
GitHub Copilot appAgent-first developmentAutomation in the cloudWorking treesMCP integrationsCI/CD and agent mergesVS Code integrationCLI parity and model strategyPrototype with multiple agentsFeedback triage across repos
Full Transcript
My wife has a couple stores. We have to track a lot of data. And so I'm using this app to manage that. And I've been able to get that to such a high bar from having agents help me out with that. Excited to be here. First time. Listen, and it will not be the last. This is the single most exciting feature for me because I've been on the inside, right? So I had a tiny little contribution when it first came out and we were all kind of like just crowd building this amazing app. And I'm super excited.
So please tell us what is the GitHub Copilot app? of the evolution of how people are using AI to build and design software. Previously, we've done a lot with CLIs and CLIs are great and I've been all in on the CLI game and I have like four to six windows running different sessions. And we wanted to simplify that entire workflow and make it to where if you're building software or you're trying to use these AI tools, let's give you a beautiful UI that makes it really easy to to do those things that you're trying to do.
And so the GitHub app just exposes quick chats as well as repositories for your source code or for even project management type things. And we try to make it to where it's easy for you to get going. So we're able to, yeah, just take an issue, click on it, start a session. And from there, it will actually go through and do the work and and fix the code. And I can go and I can view the code changes if I'd like. And if I trust it, I can run it locally or I can just say create a PR.
And at this point, the agent's going to deal with creating a PR and actually pushing the changes. And I can even ask it to monitor that PR so that way when the CI passes and everything else is ready, it will automatically merge that PR in for me. No more sitting there baby sitting CI, which is really nice. So we do have enable agent merge. Now that it created the PR, I can actually go up and click on this and tell it to do agent merge, to which now the agent will actually keep track of that and make sure that any code reviews that are done and everything that it'll go through and actually approve and merge that PR as soon as it's ready.
One of the other things that I really like about the GitHub Copilot app is we have the concept of automations. And earlier you mentioned cloud. Um it's great that like I have triage running for this GitHub app running in the cloud. And it'll actually go through and triage and run our issues. And it does this every morning for me. So when I get up and get to my desk, I can come and see which issues that we should be looking at and what we should be working on to help us make sure that we're actually dealing with this.
And that automation is actually running on the cloud every single morning. It's nice to be able to just schedule these prompts or certain things. I have it working on my work IQ. It goes through and checks my emails for me. Checks my my teams. So I actually have it transcribe my notes from my meetings and come up with my customer notes for me. So being able to have these these tie-ins with MCPs and automating the stuff, I can make sure that I'm staying on top of my work. Not even just my code, but my actual PM management as well, which is really nice.
I mean, this is blowing my mind because now this is a different level of agent workflow here. Like I don't even have to set up cron jobs to things for me. It's automatically done in the Copilot app. I don't even have to bother. This is awesome. Yeah, absolutely. We have a couple things that are built in to give you some ideas to get going like performance improvements, um triaging issues. Any of your skills that your repos pick up, you can actually automate those skills running automatically as well. And you can just go in, tell it which repo you want, when to run it, and it'll automatically run.
And you can choose to run it in the cloud or you can run it locally. It would be great to just have that happen and not deal with crons. You can even do a quick chat and have an agent set those up for you. You could say, "Hey, make sure you run the skill every day at 9:00 a.m." and from that quick chat, it'll set that up and and do that for you. I even use quick chats to set up my MCPs because I don't want to have to deal with editing JSON or making sure I have the right version of node installed and everything need for those MCP servers.
So, having this this quick chat available, just jump in, have it set up my machine so it can do what it needs to is really empowering. Tell us some of the use cases. Like, what are you using it for? What have you seen people in the team use it for? This app is a big app for us to build and our team is I think we're up to like 10 different people working on it now and now that we've opened it up when we went through our private preview stage, we had a wait list and I actually used this app to manage the wait list.
I didn't want to It's a lot of work to go through and figure out from you guys signing up, sending us a survey, we get an Excel sheet, turning that into let's link the GitHub handles that were passed in with subscriptions. We want to make sure you have a subscription that can use it. Um our business and enterprise were already had access. So, we don't even want them to do wait list. We just want to send them email. Hey, you're in. And automating that workflow, being able to use an agent, I just op- created a new folder, new repository, just said, "Hey, here's the Excel sheet where it comes in.
Here are the skills I need. Help me build those skills." And having the agent go through and build those skills. So, now I just say, "Hey, we have a new import the latest um wait list." And now I can just talk to it and say, "Hey, how many people do we have using these subscriptions and things like this? And let's invite 100 people across these different subscriptions." And it's able to generate that list and keep track of that. So, I'm not keeping of importing this, managing, and tying things in, all that together. Our team does use this also to manage the GitHub app.
I'm using it to manage all of our feedback. Being able to have the agent go check each of those repos, each of the teams' channels or Slack channels, look for feedback, file an issue, make sure all the issues are on the right repo so our team sees it, and then help us to elevate which ones are most impactful, which ones are being duplicated across them. Having it deal with that feedback and the load it takes to watch three repos, 10 channels, that's more than anyone can handle. And having an app deal with that and have it build the skills so I can say, "Hey, go refresh our repo." And off it goes is amazing.
And so now you have visibility and the power to do things on the go, whereas before, like that sound like it would have been a lot of work for a single person to manage all of that infrastructure. Uh and the Copilot app does it for you now. You're also actively using the app for the development of the app. Absolutely. Um I have feature branches and I'm working on different prototypes. And I'm doing all of that directly inside the app. I Luckily, it does things based off of working trees by default, which is great. Um not everyone's familiar with working trees, um but what working trees is, and when you're doing agent-first development, it's very helpful because when you create a working tree, what it does is it'll create a copy of the source code that only that agent's going to work on.
So each of these sessions, they're each working on a different folder and a different copy of that source code, all based off that branch, whichever is your default branch on GitHub. And now I can have multiple agents work on different things. So some of my prototypes that I'm working on, I have inside the app and I'm able to do that without any addition. I'm working on how do I add work um IQ as a default MCP integration and I'm able to I'm doing that using the GitHub app. And I'm able to then run and test it right inside the app.
And as we're all going through this workflow of like becoming closer to using agents, I find myself using reviewing the code less. I'm more worried about I course do worry about code and quality, but I'm more worried about the outcome and did it actually build what I wanted it to? Being able to just come into this app, click run, it's going to launch a new instance of the Copilot app, and now in there I can actually see that change happening and I can test to see if it built what I wanted and if that interaction's correct.
And if not, I can now give a feedback. Hey, I don't like your button placement or this doesn't work right or scrolling doesn't work like it should. And I'm able to really now focus on what the user would see and that UX experience compared to just the code. And being able to do that all in one place is a lot easier than having to context switch. So, if I'm jumping to the GitHub website to go look for the next issue and then or I'm reviewing PRs or I'm working on this. Where here I can just say, okay, here's an issue, start it, run it.
Go grab a PR, get it started, come back. When that issue's done, click run, test it, play with it. All without me having to think about getting things into the right state, having multiple instances checked out or things working. It's just all there and working exactly how I want to work in one app. Okay, I love that. But also, I see that we have the option to open what we're working on on our editor as well. Yeah, you can choose to open it into VS Code or Xcode if that's what you're working on or Visual Studio's an option as well.
You can open it into the environment you want to to go and use it the way you want. I just find myself doing that less and less the more I get used to this style workflow because if I have it set up and I just click run and it just does it. Yeah, it just does it. What models am I able to use with the Copilot app? So, the Copilot app is actually built on the CLI. Each of these sessions is a new CLI session. And so, we expose the same exact models that you would have through that.
Um, sometimes some of our features are a little bit lag behind because we need to expose those, but anything you can do in the CLI, you can do here. Very interesting. So, I take it eventually we'll be able to bring our own model then. Absolutely. That's something that we will be adding in the future. And there's lots of interesting use case with local models or when you start using the frontier models working with local models. Um you can do some really interesting use case and really drop your token usage. So, those are things that we're definitely interested in providing for everyone.
I can't wait to see how you do that cuz if you make that easier for me than trying to set it up on the CLI, I'll be a power user of both. Amazing. Yeah, definitely. Anything you're having issues with on the CLI? Um I just open up a quick chat and I tell it to set that stuff up for me. So, I encourage you to try that inside the GitHub app. Just hey, set this up for me. Anything you had previously set up in the CLI should show up and work as well in the GitHub app.
This app has changed so much from the first time that I dug in and did my little baby PR and now it's evolved with the possibility to extend into even our partners. Like there's a lot happening. What would you like to see in the future of the app? What I want it to do is I want to just have a chat. I don't want it to code. I don't want it to do anything like that, but I want it to know like help me manage this project. Help me deal with issues. Help me know what's next.
And it can be like, "Hey, I can just jump in and say, what's the status?" And it could be like, "Okay, your daily active users is this." It's tied to analytics. And it shows me a chart on the side in the canvas. And it tells me here are the top issues. And really have it help with that type of thing, not just make it like, "Okay, I got an issue. I got a PR." But let's be proactive and let's come up with what to do next and help me manage that project in a better state.
It sounds like there's a bright future for the Copilot app. I'm so excited to see where it's at next year. I mean, even later on it's already changed so much. James, thank you so much for bringing us to GitHub check out. Yes, thanks for having me on. This has been great. All right, and you you folks have a discussion as well in the community. So, for friends who are just now trying it, please go by the community discussion. We'll put all the links on the show notes so that you can give the team feedback. I can attest that this team is they're cooking, but they're also listening and they want this app to be everything you want it to be as well.
So, be sure that you go by the discussion, leave feedback, uh tag you on socials. Where can people reach out to you, James? Clancy on GitHub. I'm JTClancy on most everything else. Perfect. Well, thank you, James. All right. Thank you, Andrea. And that was your first look at the GitHub Copilot app. Automation, coding, and workflow management together in one place. If you're trying it out, please check the links in the show notes and share your feedback with the team. Thanks for watching GitHub check out. Push to changes domain and I'll catch you on the next release.
More from GitHub
Get daily recaps from
GitHub
AI-powered summaries delivered to your inbox. Save hours every week while staying fully informed.



