Boost Productivity with Copilot in Visual Studio
Chapters10
Introduces the new agent preview and how it ties into the Copilot CLI SDK, with upcoming features from the CLI and VS Code that will arrive in Visual Studio.
Copilot in Visual Studio 2026 turbocharges workflow with plan, agent previews, multiple models, and automated testing and planning from meetings.
Summary
Wendy Brighting from the .NET team walks through the upcoming Visual Studio 2026 Copilot features, emphasizing how an integrated Copilot experience can plan, build, test, optimize, review, deploy, and modernize code. She highlights an agent preview that leverages the Copilot CLI SDK, mirroring experiences from VS Code like bypass and autopilot modes. The demo showcases a configurable tool picker and a skills picker, including a code-review skill that can run two model reviews in parallel for safer analysis. Brighting also demonstrates the session manager to orchestrate multiple Copilot sessions and introduces Plan mode for structured, interactive planning. A standout feature is Work IQ, which uses meeting transcripts to generate an implementation plan, eliminating the need to rewatch meetings. The talk wraps with the debugger, profiler, and modernization agents, all designed to simplify profiling, debugging, and updating older .NET apps, with mid-June availability for the Visual Studio release.
Key Takeaways
- Copilot for Visual Studio 2026 includes an Agent Preview that integrates with the Copilot CLI SDK and brings VS Code-inspired bypass and autopilot modes.
- A configurable tool picker and a skills picker let you enable/disable tools and tailor the coding review workflow, including multi-model parallel reviews for safer results.
- Plan mode in interactive Copilot helps you create a concrete implementation plan from changes and diffs before coding begins.
- Work IQ connects meeting transcripts to Copilot, generating a ready-to-implement plan from team discussions without rewatching meetings.
- The Test Agent can generate targeted unit tests and assess code coverage for a specific scope (e.g., web app project), then run builds and tests automatically to ensure a clean state.
- Mid-June release window brings these features to stable Visual Studio, expanding Copilot capabilities across debugging, profiling, and modernization.
Who Is This For?
Essential viewing for .NET developers and product teams using Visual Studio who want to accelerate feature planning, automated code reviews, and test generation with Copilot. Great for teams adopting Work IQ for meeting-driven planning and for those exploring Plan mode and multi-model reviews.
Notable Quotes
"Hello. Um, I'm Wendy Brighting. I am a product manager on the .NET team and today I am excited to talk to you about the cool things that co uh co-pilot in Visual Studio 2026 provides."
—Introduction to Wendy Brighting and the topic of Copilot in VS 2026.
"One slide, I'm going to dive right into Visual Studio and start showing off a lot of the really great things that we have."
—Sets up the demo focus on what’s new in VS Copilot.
"The model picker here provides you with lots of really great information about which models you can use."
—Pointing out model selection and cost/credits awareness.
"The test agent uh helps you to get your code up to a uh more code coverage than what you uh have currently."
—Introducing the Test Agent and its purpose for improving coverage.
"I love the fact that it thought about the tests that needed to be added that go along with that change."
—Appreciation for Test Agent generating necessary tests alongside changes.
Questions This Video Answers
- How does Copilot in Visual Studio 2026 handle code reviews with parallel models?
- What is Plan mode in Visual Studio Copilot and how does it work?
- How does Work IQ turn team meeting transcripts into actionable development plans?
- What improvements do the Agent Preview and CLI SDK bring to Visual Studio Copilot?
- When will Visual Studio 2026 Copilot features be generally available and how can I try them early?
GitHub CopilotVisual Studio 2026Agent PreviewPlan modeWork IQTest AgentDebugger agentProfiler agentModernize agentBYOK and model options
Full Transcript
Hello. Um, I'm Wendy Brighting. I am a product manager on the .NET team and today I am excited to talk to you about the cool things that co uh co-pilot in Visual Studio 2026 provides. Visual Studio 2026 works with you to enhance your workflow. So whether you are planning, building, testing, optimizing, reviewing, deploying, or even modernizing your code, C-pilot in Visual Studio 2026 provides you with an extra set of tools that makes it easier and faster for you to get what you need to get done done. One slide, I'm going to dive right into Visual Studio and start showing off a lot of the really great things that we have.
Now I am in an internal version of a preview version of Visual Studio. So these are all things that are coming uh within the next month um in the June release of Visual Studio. So um if you have the the stable release of Visual Studio, some of these things are not available quite yet, but they are coming in the very next release. So I wanted to really show off a lot of what's what's what's up and new and coming to make your life even easier if you're already using Copilot. One of the things I wanted to point out here is this agent preview.
So you're probably using agent mode quite regularly, but we are adding agent preview. This is going to be our new harness that is going to take advantage of all of this uh copilot CLI SDK that is being developed. So, all of those features that you saw in uh the CLI earlier this morning are now going to be coming to Visual Studio as soon as they hit the CLI SDK, which is amazing. Lots of really great stuff. Also, you just saw in VS Code their uh bypass and uh autopilot mode. These are also coming to Visual Studio, so you'll be able to have all of that great feature as well.
Um the model picker here provides you with lots of really great information about which models you can use. There's going to be um notifications on the um cost, how many credits that they will use. I'm going to stick with GPT54 today. Um but most of the stuff that I'm working on is going to be able to be used in across any of your favorite models, whatever whatever you prefer in including the bring your own key uh stuff that was shown in VS Code. also going to be available in Visual Studio. Some of the really cool things, the first thing I'm going to start, uh, this is our tool picker.
Lots of changes where you can turn on and off which tools you want to have available down to even which ones you want to include. We also have a skills picker. We just heard a lot about the skills uh Matt was talking about and I know they were mentioned in the previous ones. Here you can this lists all the skills that I have installed and you can turn them on or off. You can also come in here and edit them directly. So it brings up the MD file that describes this um this skill that you're asking to.
This one is the one I'm going to use right now. Um here this skill walks it through on how I like it to code review. So yes, we have lots of great code review options. I have I'm very picky about how my code reviews should be done. I walk it through the steps that I want it to follow, how I want it to present the results at the end, giving it the format that I want to see things in. And most importantly, I wanted to make sure that it does things in a um a little bit safer way by doing a parallel analysis with multiple models.
So, I have included in my um at the bottom here of my skill that I wanted to jump off and fire off two separate reviews at one time and then bring that together uh with an analysis. So, my first prompt here, I'm just going to paste in. I had just finished a feature um that allows for I'm an ehop. I want the users to be able to add items to their cart even when they're not logged in. Of course, that brings in some complexity where I had to if they had a cart already in their loggedin version when they go to checkout, it needs to combine those.
So, there's it's a little bit complicated. So, I'm going to ask it to to go do that code review. And so, it's going to think for a second here. And then the next first thing we should see it do is kind of say, okay, I need to check out that code review scale. There we go. it has recognized that skill and it's reviewing it and then it's going to fire off a bunch of stuff. Now, another great thing about the new co-pilot in Visual Studio is that's off and running. I don't want to just sit here and watch it because a code review is going to take a while.
It's got to go get all the diffs and understand, run tests, make sure do all the things that you would do when you're code reviewing and I don't want to sit here and watch that and you don't want to sit here and watch that. So, we're going to go out and I'm going to show you the session manager. This shows you all of the co-pilot sessions that you've got. These are ones that I've run in the past. That's the active one that is running. I'm going to go ahead and kick off a new one. So, I've got a code review going on of my last work.
And now, so that you don't see me have to watch me type, I'm going to bring up my next prompt. For this one, I am going to use the amazing plan mode that just released in Visual Studio Copilot. So, this one, I'm co- reviewing my last change. I need to plan my next change. So, I just implemented this new feature, but I realized I didn't do it in the chat mode. And so, if a user is chatting with our AI chatbot in ESHOP and the chatbot wants to add an item to the cart, if they're not logged in logged in, it's going to not do it for them.
So, we need to get that anonymous cart adding in the chat. And so, I'm in plan mode. I'm in interactive here. Um, I could switch to bypass or or autopilot, but I'm going to keep it in interactive for now. And I'm going to fire that one off. So, that one's going to think, going to reference all the things it needs to reference and get going. Let's go check on our code review, see how that's going. Wants to access a few files here. So, let's confirm those so it can keep going. When you have multiple things going, it's probably better to switch to bypass or autopilot mode, but I'm a I'm token conscious kind of person and I like to watch what it's doing and stop it if it starts to go off the rails.
So, um I'm more of a watch it kind of person. Everybody has their way of doing things and uh and that that's great. That's why we have these options. It's amazing. So, that one's still running, doing inspecting. Um, it's now kicking off those two independent reviews in parallel, grabbing one cloud model and one GPT model so that it can synthesize across those. And those reviews have been kicked off in in parallel tasks, which is amazing. Let's go check on our plan. Lots of reviews happening. going through and it accessed all the code that I changed doing git diffs and everything from it because I have committed up to a branch that I'm working on most of the changes are not still just locally so it had to go out and look through and find and understand it's now come up with a a plan here and so looking through found that it needs to make changes to the basket state uh some changes to the typescript and some test files So, it's pretty comprehensive.
I don't think I have anything here. So, I'm going to say finalize this plan. No changes. And so, it's going to go now and create a plan document, an MD file that I can save and review or I can start implementation on. So, here it's created that plan. I can open it directly from here. really good description of the steps it needs to take and and then there's a button right there for it to start implementing. Um, I'm not going to watch you implement. You've all seen it implement good code. We don't need to go through that.
Um, before I kick off my next one, let's go check on our code review and see how that's going. Some more files it wants to have access to. All right. All right. While that is going on, the next thing I want to do is show you this really but lifechanging thing for me that I found. I'm going to start off another chat. So, how many times have you been sitting around discussing in a meeting this great feature? Amazing. You're recording the meeting. The meeting's got basically laid out all the requirements and everything you need to do for this feature.
And then you have to go back and you have to write up exactly what you all just talked about in that meeting. Instead of that, now there is this amazing thing. I'm going to switch back to this agent preview and I'm going to ask it to use my work IQ uh tool and get the transcripts from this amazing meeting that we had that planned out the changes that we need to be done and I am asking it to create the plan that uh is needed to make the changes based on that conversation. And so now I no longer have to go find the transcript or rewatch the meeting or try to remember what happened.
It's all going to go out and grab that information from the team's call. Get the transcript. And so it's here. It's getting work IQ access approved. So this the first thing it has to do is do I approve it to do this? Absolutely. Um and now it's going to start planning those changes. It's going to review through the transcript, make sure it understands everything it's going to talk about. This was a multi-session meeting. So, we've had three discussions through this meeting. So, it's going to go through the transcripts of all three of those meetings, and it is going to formulate this amazing plan on how to make the changes to this site happen.
Let's go back and check on our code review. All right. It's been doing a lot of stuff, reading, checking things, understanding the diffs, multiple agents running at the same time, trying to understand and review this code to make sure that it is going to provide me with the best information that it possibly can on how I can get this code ready for production. The two parallel code reviews are still running. It's saying and it and so it's going to continue waiting for those results. So we'll come back to our work IQ and review. It's reviewing the transcripts and coming up with that amazing plan.
Some of the things that I'm not going to have time to be able to show you that I wanted to make sure we have uh an extra I wanted to mention. Um the I said we have a lot of great things no matter what you're doing in Visual Studio. The debugger agent and the profiler agent provide you with the easiest ways. Profiler is I've heard a lot of people say profiler is so hard to understand. I don't know how to profile. I know it's important but I don't know how to use it. This profiler agent makes it so easy for you to understand what you need to do.
Profiling your code and provides you with the changes that you need to make to make your application so much better. You're running into bugs, the debugger mode will help you with your breakpoints and all of the things that you need to do to find out where the bugs are happening and what's going on there. All amazing things. And then if you're using some code that's not quite as up todate, you're not on net 10 yet. Um, modernize provides you with an easy way to get your applications up to the latest. Whether you're in .NET Framework and you need to go to .NET Core or you're just an a version or two behind, the modernized agent will really help you bring up.
It's going to provide you with a plan what it needs to do to get there so that you can feel comfortable in making that move and you can make it peace meal or all at once at your comfort level. that modernization agent provides you with lots of great features there. So, that one is still working and that one is still working. And so, I'm going to kick off one more. Oh my gosh, three at one time. Um, but you know, we can do that these days because it's not taking my brain power, it's taking the LLM's brain power.
And I'm going to show off the test agent. The test agent uh helps you to get your code up to a uh more code coverage than what you uh uh have currently. Uh making sure that it is covering the most important code that you have. Um not implementing tests on things that don't need to be tested, but on the important things. So I go into uh the test agent here and I provide it with what scope do I want it to provide coverage for. Now my changes that I made were in the web app project.
And so I'm just going to say um pound web app and I'm going to select that web app.cs. And so it's saying scope the the test generation down to that web app project. I don't need to add coverage to the whole thing. I didn't make changes to the whole thing, just that scope. And so this is going to it understands the scope and it's going to kick off and generate tests. This can be quite long running um depending on how your coverage is. But the first thing it's going to do is it's going to assess your coverage, right?
It's going to have to run your tests and do an code coverage analysis and understand where those gaps are. And then it's got to review your code so that it understands what are those important lines of code that need to have coverage, where the gaps are. And so it goes through and does all of that. Uh generates a whole bunch of tests. And I'm going to let that one run while it's going again while we go back and check on how our others are doing. And this is kind of the flow that I have going on a regular basis.
I have a code review going of my last change. I got a plan going on my next change. Um and then I am starting development on the one that I previously planned. And so lots of things all going at once. Everybody has their different styles of the way they're doing things. This one is still working on um it. How is our work IQ? Still analyzing those transcripts. Um, when you're live streaming, everything takes longer. These yesterday took me five minutes to get this plan. The code review takes a little bit longer because of the reconciliation.
That one takes about 10 15 minutes. Um, the test generation one does take a lot longer. That one took about 20 25 minutes to run. But with this one yesterday, it generated 161 tests across 21 files. It is already started. If I go over to my git changes, we can see that um it is we can watch it add tests as it goes. Um, it is currently generating unit tests for looks like four files it's identified don't have good coverage in in our output here. Continuing to work through and add things and we can go back.
Yeah. Generating unit tests for our order status changed our product image provider. all of the different things that we had some some gaps in our in our knowledge there um coverage for for the tests which amazing because who likes to write unit tests we want to create new features not write tests and this now does it for us and at any level that we need this one I'm doing at project level I could do the whole solution level um I could do it on just a file or a method uh it provides all of those uh options depending on the scope that you are looking to offer.
So, lots of great options with this test agent. As that's working, let's see how we're doing here. I'm wondering if this one got stuck. It doesn't seem to be. Sometimes Copilot gets stuck. Um especially when you got multiple things going at one time. Even when I'm using the CLI, I find that sometimes I have to give an agent a kick in the pants to get it going back again. Um, other times it's moving along. Um, and we're all good. That one seems to be still running. And here we're it's generated some tests. It's now um performing a build to make sure that it found no errors or it sees if it has anything.
See, we've got new files that were added. Um so, it's working away on making those changes. I can go back to my uh this one. I can get out of plan mode. I can ask it to implement. It's going to follow the plan specifically at how I had asked it to. Um, I could go in here and make changes, but it said this plan looks great. That's exactly the changes that I would need to make. And I love the fact that it thought about the tests that needed to be added that go along with that change.
Um, and so lots of great uh things going on there. It's finished its tests on this one. It is still and now it's added on another one. So, it's adding test more chate tests our work IQ. H I wish this one would finish. In fact, because this one's still running along, I'm going to show you what it did yesterday. So this one um it mapped out the work IQ for the um session uh for those for all of those meetings that we had to talk about what changes we needed to make. It created the plan.
I can open the plan here. It describes the problem that was discussed and how we can use um that to make this product even better. Um and so then it suggests an approach. It describes the plan steps on how it can align uh the Aspire SDK versions and remove design time, build noise, make onboarding and repo prerequisites accurate and explicit all the changes that need to be made based on the conversations that we had. Um and then I can just dive right in and say implement this plan and no extra work from meeting to making the work happen.
Um this particular work IQ tool was has been a lifesaver for me because yeah I sit as a product manager I sit in the meetings all the time and used to have to go back and then translate all of those conversations into a spec with co-pilot and work IQ. I no longer have to do that. It's created that spec for me. I can have the team review this before implementing, making sure that we're all aligned and that it got everything correct as we discussed in the meeting. And just I can send it off implementing which is absolutely amazing game changer.
Um from the uh code review perspective since that one is not Finn and Shang I can do the um the uh baking show change here and show it went through and it reviewed with two different models and it is providing me in the format that I have asked it to um what changes need to be made before I should feel comfortable with moving this change into production. So it has outlined there are some reliability issues that I might want to look into with regard it's pointed out the lines um that are of code that are impacted and what is the recommended fix.
It I had it look at very specific things around thread safety and security and test coverage things that are all very important to me. It says from those perspectives those are good. It puts a green check mark. Um then there are some suggestions. These aren't absolutely necessary, but I suggest you look at these um when you're looking at, you know, clearing browser carts and all of that. There there's some possib some improvements there. And then it summarizes all of that into here are your critical fixes that you need to make. There is a little test gap which I'm covering with my test generation and um the recommended changes that we need to make before merge.
a really easy, great code review before I even send it out to my uh the rest of my team to code review. Keep things a lot less burdensome on the team. Makes their code reviews a lot simpler because I have taken the time to do this multi-step code review um for myself. I can make those changes. It makes my code a lot better before I go out and ask my team to spend time looking and reviewing through it, which I love because nobody wants their time wasted. all right. And so it has finished making those changes to the chat app to do the based on the plan.
All of that it's got, you know, two files that it updated plus that plan. All of that is complete. My tests are still working. If I look over at my git changes, our list of added files is growing. We've now got five added files. As I said, um when this completed yesterday, it was up to I think 14 uh files added. So, uh lots of good. It will do a test run. Um, it builds and tests everything to make sure that everything it's leaving you in a good state with all passing tasks before it finishes.
Um, so amazing test agent. And that that is all I've got for today. Um, these are going to keep running in the background. I'm on a dev box so they can keep running as I wrap things up. Um, which is also an amazing thing if you're not on a dev box. This allows for me to shut my laptop and keep things going in the background there and everything is is fabulous. These features, the uh new CLI agent, the um bypass and autopilot mode, all of these things um are coming in uh mid June in Visual Studio.
Um and so thank you for watching and back to you, Matt.
More from GitHub
Get daily recaps from
GitHub
AI-powered summaries delivered to your inbox. Save hours every week while staying fully informed.









