How to ship from issue to merge with the GitHub Copilot app | demo

GitHub| 00:05:00|Jun 2, 2026

Chapters12

Introduces the problem of context switching and shows how the Copilot app consolidates issues, PRs, and sessions in one place.

GitHub Copilot app unifies issues, PRs, and automation, letting you plan, implement, and merge across multiple sessions without losing track.

Summary

GitHub’s Copilot app is shown as a single-pane workspace for developers juggling multiple issues and pull requests. GitHub’s demo uses a coffee shop homepage to illustrate starting sessions for five items, each in its own work tree to prevent cross-branch bleed. Sessions stay isolated because each item gets its own folder on disk, so editing in one tab won’t affect others. The presenter highlights staying outcome-focused, actively providing feedback to the model in plain English rather than “set it and forget it.” As work progresses, diffs can be reviewed at a high level, and changes are deployed from the work tree by running the workspace or launching a server locally. When the PRs require attention, Agent Merge continuously checks review threads, CI status, and branch cleanliness, automatically landing changes when appropriate. Automations cover repetitive triaging and cleanup tasks, which can be tuned with a simple click. The demo culminates in a streamlined SDLC where triage, planning, implementation, reviews, and merges happen largely within the Copilot app, dramatically reducing context switching and manual babysitting of PRs.

Key Takeaways

Each issue or PR can be opened in its own Git work tree, isolating changes so editing one item never touches another.
Agent Merge monitors review threads, CI status, and branch cleanliness, and can land PRs automatically when everything is green.
Sessions are created per item and can be managed with simple “new session” + “start session” actions for streamlined workflows.
The Copilot app can generate, review, and iterate on a plan in plain English, then implement within the work tree and let you review high-level diffs.
Automations for triaging issues and sweeping stale branches can be enabled with a click, speeding up repetitive SDLC tasks.
The workflow emphasizes staying outcome-oriented, feedback in natural language, and relying on the model to implement while you guide.
Ship rates improve dramatically as shown: “three of them ship themselves” compared to pre-demo days.

Who Is This For?

Software teams and developers who want to reduce context switching between issues, PRs, and reviews, and who are curious about using Copilot to manage the full SDLC from a single app.

Notable Quotes

"Every time I switch between them, there's a cost, a new window, a new branch, and the effort of remembering where I left off."

—Sets up the motivation for why the Copilot app's unified workspace is valuable.

"The GitHub Copilot app keeps it all in one place."

—Announces the core benefit of the demo setup.

"I'm staying outcome focused, agreeing the plan, giving feedback in plain English, and leaving the implementation to the model."

—Highlights the collaborative, human-in-the-loop approach with the agent.

"Agent merge is what makes the multitasking actually pay off."

—Emphasizes the role of automated merging in handling multiple streams.

"Three of them ship themselves and two issues already have a head start."

—Shows the dramatic efficiency gain demonstrated in the video.

Questions This Video Answers

How does the GitHub Copilot app manage multiple PRs and issues in parallel?
What are work trees in Git and how does the Copilot app use them for isolation?
Can Agent Merge automatically merge PRs when CI passes and reviews are complete?
What is the workflow for using Copilot to plan, implement, and ship changes from start to finish?
How can I customize automations for triaging issues and stale branches in the Copilot app?

GitHub Copilot appAgent MergeWork treesCI automationAutomationsSoftware development workflowSDLC automation

Full Transcript

Most days I'm juggling a lot. There's usually a bug to investigate, a couple of pull requests to review, and a fact I started yesterday, but didn't really finish. Every time I switch between them, there's a cost, a new window, a new branch, and the effort of remembering where I left off. The GitHub Copilot app keeps it all in one place. Let me show you how it works. For my demo, I'm going to work on my coffee shop's homepage. I'll open the GitHub Copilot app and start in my work. It brings all my issues and pull requests from every project into one view. They're grouped into sections for quick scanning. Active is items assigned to me or opened by me. Review requested are items waiting for my review. And done is the cleanup section. I can also add custom sections. I just click on the plus and a section can be anything I can write as a GitHub search. So today I have three pull requests and two issues to work through. Now for each I need to start a session. So I can click on one and then select new session. So I'm going to do this for all of the five items. skip a little bit ahead and come back to it with all of the items already started. Now that all five sessions are open, let me show you how they're isolated. For each item, the app checks that branch out into its own folder on disk. It's called a work tree, and each session gets its own. Editing a file in one tab doesn't touch the others. No stashing, no get checkout shuffle, no risk of bleeding changes from one branch into another. Looking at the dark mode toggle issue, we can see that the agent already has started on the issue. This is because I enabled auto start issue sessions on the project. The agent has already drafted a plan and committed to the work tree. The plan looks reasonable, but I want the toggle in line with the page title, not in the header. I'm going to tell the agent that and let it implement. This is the part I want to be clear about. I'm not just pressing start session and walking away. I'm staying outcome focused, agreeing the plan, giving feedback in plain English, and leaving the implementation to the model. I can review the diffs if I need to, but I want to stay high level. Once the agent is done, I can click run at the top of the workspace or start a server via the terminal. The site comes up in localhost straight from this work tree. So I'm seeing exactly what the agent built. If I want to change something, I use pick and polish. I click on an element and a screenshot of the surrounding context attaches to the composer. No copy pasting selectors, no describing what I'm pointing at. Now let me come back to a pull request. The stylesheet PR has a co-pilot code review enabled and the CI just failed. Normally, this is where I would start context switching, read the review comments, dig into the CI log, push a fix, wait for CI again with all three PRs cued behind me. That's most of my day gone. Now, this is where agent merge handles it. So, every few minutes, it checks three things. Have the review threads been handled? Is the CI green? Is the branch clean? If not, it works on them. If it gets stuck, it pings me. Otherwise, it lands the pull requests on its own. With me multitasking myself more than ever, I can't spend all the time babysitting PRs through to merge. Agent merge is what makes the multitasking actually pay off. Next, let's look at automations. So, these are the part of the day that I do over and over again, like triaging new issues or sweeping stale branches. And I won't dive into here, but they're just a click away. Well, that concludes the demo. This demo showed that the entire SDLC can now be done from the app. I can find the right work to focus on. I can partner with the model to build a plan, implement that plan, stay outcome orientated in my feedback and land changes with agent merge to close out the entire SDLC. A week ago, a good day for me was shipping one pull request. Today, three of them ship themselves and two issues already have a head start. That's it. No tab juggling, no stashed branches, no losing my place. Whether I'm triaging an issue, drafting a pull request, or even working through review feedback, the agent works alongside me and ships the work.