Transcription
There were tons of announcements at GitHub Universe or on GitHub, GitHub Copilot, Visual Studio Code, the Agent HQ, and a whole bunch more. Today, I want to talk about some of the biggest ones coming to VS Code that I've gotten a lot of questions about, which is the agent sessions panel and the new OpenAI codec extension that's been updated to enable you to actually sign in with your GitHub copout subscription. It shows up next to all of your agent sessions and you can actually get builds to your GitHub copout subscription as well. I want to talk about why you might want to do that and what that experience looks like.
There are tons of other announcements that you probably want to dive through, so definitely check those out. Um, I'll put a link to the full release notes. I'll put a link to the uh keynote video and the keynote kind of like here's what the announcements were. You definitely want to check this out. It's awesome. But I'm going to talk about these first two, which are the agent sessions and the codec stuff.
Okay, so here I am over in VS Code and this is sort of our normal view, right? So here we have our chat on the right. We have our explorer on the left. Of course, you can, you know, move these around if you want to, whatever you want to do. And here, I've just been asking it a little bit of information about this MCP badge creator project I've been working on. And, you know, this is looking really nice and just kind of standard. Now, you can always come in and you could look at all the history, for example, of your chats. Go over here. You could come in and you could say new chat, which would add a new chat here in the pane. You can add a new chat editor. So, that'll actually open up a new chat over here. You can actually open up a new chat in a chat window. You can run all of these simultaneously, right? Uh right here. So you can have an ask an edit session, different agents working. You could have playright running for example. You could be working on different front-end components all at the same time. Like they all work all at the same time. You have as many agents working here asking going deeper on projects. If you're working on a back end and a front end, you can have those all working all at the same time right here. So I can come in, see all these chats. I can also come to this right this little chat little icon up here with the sparkles and open and close that main chat on the right hand side. I can also toggle it over here. I can come in, I can open the chat, I can open the chat editors here. So I have a lot of control going on here just just like you would from the main chat as well. So if you don't want to have this main chat open up all the time, you might want to have a chat that's pinned for example. You can do that as well. You can do use different agents, you can use different models, you can use whatever you want.
Okay. So, let's go ahead and see what's new here. Because if I have tons of chats and tons of files open, it can get a little unruly. So, over here on the lefthand side, we'll see this new agent sessions panel. And what this is going to show me is all of my agent sessions. Ones that are running locally, ones that are running in the cloud, or ones that are even running in the CLI. Okay, so that's really, really cool from a high level. Just everything in one sort of little mission control here. And I can tap around and I can open these different chats, right? Right. I can even close all the chats here and they all go away. I can come in and open new chats. Right. So I can go in and kind of open these. I can come in, open it in a new window side by side. I can kind of do anything that I would normally do, right? So if I came in, I could ask it something. It would get to work. So really easy to be able to manage it. You can rename them. So your chat history is all here as well, which is cool.
The other thing that you can do is you can actually start a cloud agent session. So this is the get a coding co-pilot coding agent. Uh, now we're calling it the GitHub copilot cloud agent. And this actually going to delegate a task to the cloud. So I could say here I have a default agent. I have different agents like my playright agents that are here. So I could come in and say, "Hey, can you add a new orange theme to the app, right? So I'm just going to go in, send this off, and it's going to delegate that to the cloud agent." And I see it here. So this is actually going to open up a pull request for me that I can see. And this is not happening on my machine. It's happening in the cloud. So, I could actually go tap on this. Um, I could actually go dive into the pull request that was being opened for me. And then I could see it get to work at any time. I could actually say, "Oh, yeah. I don't want to actually be orange. I want it to be, you know, Christmas themed, you know, or or Hanukkah themed, for example, right? And I could go in and I could actually delegate different items. I could stop the session all from here. And we can see it all spinning up here as I could add contacts. I could add files or folders. So that's really nice to be able to have this mission control all working here too. So that's really really cool. We're going to close that down. I could of course open that at any time and it's going to load that session up for me.
I could then also come down into my copilot CLI. So it's actually a CLI running local background tasks. So if I just want things to work in the background, for example, like do a deep do a deep analysis of security uh on this app. Right? So it's going to go off and it's actually going to delegate to the CLI agent that's running just randomly in the background on my machine just as if I had open up terminal for example and I can do the same thing. I can use different models. I can interact with it. I can add context and I could go ahead and stop it. So this is also happening on this local branch too. And we can see it's going to run commands and it's going to ask me for permission just like if I was in the command line. So again, this is just running locally, but it's not actually, you know, modifying, you know, my space here as far as like taking up my actual chat context, right? So if you like the interaction of how the CLI works, then boom, you're good to go. And I could go ahead and delegate that off. I could always see down here on the bottom right like actually things that are running. So we can see I have a cloud uh code agent session here running and doing a bunch of stuff, which is really cool. So we can tap on those, get back and forth, anything that we want. And we can see now all of them over here getting to work. So I see all my chat agents whether they're running locally or remotely, anything like that at all, which is really really cool.
Well, let's talk now about our actual uh OpenAI codeex extension. So if you go over and you type in codeex here, you might see it actually like in the agents panel down here as like a little popup, but you can actually install the OpenAI codeex extension. Now, this is an extension from OpenAI that has Codeex and Codeex has a CLI, but it also has the VS code integration and a panel. So, you can chat, edit, preview changes, work. So, if you specifically like how the OpenAI codeex extension and how OpenAI codeex works, then maybe you would be interested in using this inside of VS Code. Now, I do want to point out though that you of course also have access to GPT5 codecs and GPT5 directly inside of your GitHub copilot subscription through the normal chat. So, you can go ahead and use those as well. But, of course, you're using the GitHub Copilot experience over here, whereas you might be using Codex already and you like that experience.
So, let's install the pre-release. That's pretty simple there. And what I'm now going to do is go over and I could see that one we have a codec sort of uh panel that's been open up here but I can also go into my agent sessions which is really cool and I now see the OpenAI codecs right here. So I can see all the sessions for the OpenAI codeex side by side. I can click on it. I see a new session's been started and now I have the option to sign in with chat GPT sign in with GitHub copilot or use an API key. Now, I don't have a chatbt account and I don't have an API key, but I do have a GitHub copilot subscription. So, I'm going to go ahead and tap on that. And this is going to allow me to sign in. So, we can see here that it decides, you know, you decide how much autonomy you want from the agents. The codecs can make mistakes obviously. So, review the code. But this one is powered by GitHub Copilot. So, it uses your C-pilot plan for all model calls, billing, and rate limits. The codec extension usage blah blah blah. You have the open codex terms of service and the GitHub terms of service there. So you're being build through your GitHub copilot subscription in this case and that's where the model calls, billing and rates are coming from but I'm going to get the actual codeex experience here.
All right. So here I can obviously only set this up in a local environment because I have that here locally. I have agent chat and I have agent full access mode and I have access to the GBT models here that OpenAI is giving me access to that work with codecs. Right? So they're going to have different system prompts. We're going to have different work there. I can see history. I can look at settings. So I could actually like go into my codec settings, the IDE integration, so how I want this to open up. I could configure MCP for example. I could look at my keyboard shortcuts and a lot more. So I could say, give me a quick overview of this application and decide if there are any quick features to add. And let's go ahead and put this just into the chat mode and then hit go. So now I'm actually going to go ahead and get the OpenAI codeex look and feel here because this is using the OpenAI codeex extension, but it's being built of course through my actual GitHub copilot subscription. So I can see it put a little task list here. It's reading all the files just like my chat stuff would work, right? So I could actually still open my chat, do my whole thing, anything that I want here and have the OpenAI codeex integration side by side this here, but I get to see my actual sessions that are being logged here as well as this is sort of thinking. So this is really really neat from the fact that I have all of these different panels and chats and things open right here. So this is going to do its thinking. It's going to go ahead and summarize the current functionality, identify potential quick fixes or add additional enhancements here. And I'm actually really interested to see what happens. So, uh, which is nice and very similar right here. So, I could add context, I could add images, I could add files. So, that gives you a nice little experience.
Now, something a little bit different is this is using a premium request because it's going through my GitHub copilot subscription. But I can see here this is the context window. So, this is specifically how OpenAI and Codex works is it shows you the context window. Now, this doesn't have anything to do with billing. It's just the context that is being sent back to the model. So, I can see if my chat is getting too big and I maybe want to start a new thread there. So, that's something that I might want to think about as well. So, let's see. It's going through and it's looking and it's identifying some potent potential quick fixes for me and enhancements. And yeah, let's see what it does. There we go. It's doing a quick list and I could drop this down so I can actually see quick gathering badges text. And this is kind of a nice little view here that you might, you know, might appreciate. So here we go. So we get an overview. We get a home hub MCP quick wins wire the persistent state badge text download MCP JSON. That's kind of cool too. Provide this here. Right. And then I could go and I could ask it to then implement this putting it into agent mode for example. So I can see that the change to agent mode there as well. So this gives me this complete overview of everything happening in real time. So I have that remote agent that's going off. It's doing testing here. It actually has like different images that are coming up. I could then go into my CLI back into my codeex integration. I can see them all working here side by side. So as I spin up new copilot agents on github.com or from the modable app, they're going to show up here from the cloud agent CLI on my local machine. I can spin up new threads for example that I want to work on. And then I can go in also just, you know, work on code. If I want to go in and just open a file, work on it. I can also do that too. you know, you it's up to you whatever you want to go ahead and do and and work with as well.
So, that's sort of like a complete overview that I wanted to show of how this actually works and how you can start to use not only the brand new agent sessions pane over here to see everything working harmoniously together, but also additionally all of the copilot CLI and that integration into open eye codeex extension as well. You know, people may say, well, why do you want to maybe use all these and have all these available? Well, use what makes sense for you. Some folks like exactly how the GitHub copilot CLI works. Other folks like how the OpenAI codeex extension works. Me, I like how the built-in chat works. That's just me. But I love using cloud agent. So, having these in a dashboard, so you're in control, right? You decide if what extensions you want, what integrations you want, and you're totally good to go.
So, one other thing I want to actually talk about here is something that's pretty cool is that you can now come in and let me go and start a new chat. I'm going to say into agent mode here. I could say, "Hey, let's add a holiday themed to the app." Now, I could decide to fire this off with agent mode directly here, or I could actually hit this new delegate to agent. Now, what's cool here is because all my agents are together here, I'm in control of how I want to delegate it. So, I could delegate it directly inside of the agent mode that's running here, or I could decide to run it in the cloud agent, the CLI agent, or the OpenAI codeex. So, this is super duper powerful to be able to actually go off and delegate it to the specific agent that I want all from a single chat. I love how the team has integrated this stuff together to really make it seamless. Whether you're over here in the agent sessions monitoring and reviewing everything in real time or you're off just using the chat mode here as well and delegating it off to the agent that you want.
Anyways, I hope that you enjoyed this video. If you did, if you have any questions, put them in the uh comments below. Super appreciate that. And of course, like, subscribe, do all those things. Don't forget to subscribe so you stay up to date every time I put out a video right here on the channel. So until next time, happy coding. [Music]