GitHub Universe 2025 Day 1 Announcements  ## Metadata - Author: [[YouTube]] - Full Title: GitHub Universe 2025 Day 1 Announcements - Category: #articles - Summary: Missed the GitHub Universe Day 1 keynote? Catch the biggest announcements here! We introduced Agent HQ, transforming GitHub into an open ecosystem for all coding agents (from OpenAI, Google, Anthropic, xAI & more). Plus, get the details on Mission Control, Plan Mode, Custom Agents, and new enterprise controls. #GitHubUniverse #GitHubUniverse2025 #GitHub Watch more videos from GitHub Universe 2025: https://www.youtube.com/watch?v=P6Va0_KILi4&list=PL0lo9MOBetEFKNlPHNouEmVeYeyoyGTXC Stay up-to-date on all things GitHub by connecting with us: YouTube: https://gh.io/subgithub Blog: https://github.blog X: https://twitter.com/github LinkedIn: https://linkedin.com/company/github Insider newsletter: https://resources.github.com/newsletter/ Instagram: https://www.instagram.com/github TikTok: https://www.tiktok.com/@github About GitHub It’s where over 100 million developers create, share, and ship the best code possible. It’s a place for anyone, from anywhere, to build anything—it’s where the world builds software. https://github.com - URL: https://www.youtube.com/watch?v=nSwj2Ma0pnk [Music] We've always been the home for developers and that's not changing. Now we're adding agents with developers and agents working together in agent HQ. GitHub Copilot is now a fullyfledged coding agent far beyond autocomplete. It now operates with deeper context, stronger reasoning, and the ability to execute and coordinate tasks across the developer workflow. Why not also have Claude and Codeex in your Agent HQ as well? Or what about coding agents from Google Cognition or XAI? That's right, agent HQ is welcoming every coding agent to GitHub. So starting with these coding agents and coming later this year, we're partnering with Frontier Labs and startups to bring their coding agents onto the GitHub platform. Agent HQ offers a single place to start, steer, and manage sessions from all of them across the web, IDE, CLI, and even mobile. It's your favorite coding agent's favorite platform and it's all available with a co-pilot paid subscription. Use any of the agent HQ coding agents coming later this year to help you manage and take advantage of all of these coding agents. We've extracted core functionality from co-pilot's coding agent to provide a new platform to build on top of new branch controls so you control when agent generated code is run. Each agent has a full real identity on GitHub, meaning you can give it access to only what you want, assign it issues and pull requests, and even mention it in comments. Each agent can build on top of our existing compute layer with GitHub actions. And you get enterprisegrade AI controls to manage access, set security policies, and audit agent activity on GitHub. Agent HQ is a single command center across every coding agent you use. From a new single view of mission control, you can see all of the sessions from your agents across your projects. You can assign new tasks, track progress, and review the code as it's written. You can even spin up new sessions wherever you choose to work. You can move from the CLI to the web, the web to the IDE and work wherever it makes sense for you. We're building a single way to work with coding agents, giving them the right context about you and your work and most importantly giving you control. You can connect to core services like Figma, Sentry or Atlassian Jira via their MCP servers and the GitHub MCP registry. Your co-pilot coding agent runs right on GitHub actions, giving the agents their own hosted development environment to build and test changes before asking for your review. But for some of you, we've heard loud and clear that's not quite enough. So starting today, Copilot coding agent is able to run on your self-hosted runners, the same ones you use for CI/CD. And GitHub Copilot must have all the most powerful features of GitHub. So, we've integrated code security with CodeQL, dependency scanning, and secret protection as tools C-Pilot can run on every task. Speaking of tasks, we're also introducing custom agents, so you can tailor Copilot for specific scenarios like writing documentation or doing test-driven development. But I have another surprise for you. Wouldn't it make sense for you to use the most powerful agents where you spend the most time in your editor? Access to codeex is available in preview on VS Code insiders today for Copilot Pro Plus users and all the agents from all these partners will be available in agent HQ later this year with more details coming soon. Now on to our flagship editor for GitHub copilot VS Code. In the last six months, we've been working together with the community. We've shipped nearly 200 improvements to Cop-lot in VS Code. Full MCP support, letting Copilot understand your entire stack and not just the files in front of you. Custom rules where you define the standards you want C-Pilot to follow when generating your code. Slash commands to share common prompts and repeatable workflows across your team. and enhanced semantic code models to give C-Pilot deeper context and a smarter grasp of your code stored on GitHub. Today, we now ship new models to C-Pilot customers as soon as the Frontier Labs released them. Models like Claud Sonnet 45 and GPT5 codecs shipped the same day. VS Code extensions such as HuggingFace and Azure AI Foundry can further expand your choices, giving you the ability to use other model providers outside of your co-pilot plan. You can connect to hosted model providers like Open Router or run local models through tools like LM Studio. And sometimes all that choice can be overwhelming. And we now have auto mode where Copilot will choose what it thinks is the best model for your task. These models aren't only about features. They're about how development feels. And today, we're releasing a new version of VS Code and GitHub Copilot. And I want to start by talking about one of my favorite new features, plan mode. The most successful teams use C-pilot to create something akin to a requirements document before getting started. But now, plan mode lets you work with C-Pilot to build a stepbystep approach for your task. asking you clarifying questions along the way. This quickly improves what C-pilot can do by helping you find gaps, missing decisions, or project deficiencies early in the process before any code is written. Once approved, it takes your plan and gives it to co-pilot to start implementing, whether that's locally in VS Code or using one of your agents in agent HQ. Starting this week, you can at@mention GitHub in Slack to kick off an agent task. Just tag GitHub in a thread with colleagues and ask it to fix a bug or spark the idea that you've been working on and co-pilot will get started. And it's not just Slack. You can already collaborate with C-pilot in Teams or kickoff tasks from Azure Boards, Raycast, and starting today with linear. We launched co-pilot code review as our first pass, giving you meaningful reviews of pull requests, but it simply wasn't enough. So, you can use the new agentic co-pilot code review right now. And for our enterprise admins in the room, you can turn it on today in your enterprise by enabling it within your preview features. We could solve every single problem in software development. We haven't yet, but it won't matter if developers at work can't use the best parts of GitHub. This means making sure your boss or your boss's boss and the governance teams have the tools and controls they need to introduce C-Pilot into your company, all while showing the ROI. And we've done that in three really big ways. The first is code quality. It's no good shipping more code if that code isn't great. Quality is more than just a passing review because a core problem is that looks good doesn't mean is healthy. Otherwise, we'd never have any bugs found in production or live site incidents. And that's where GitHub code quality comes in. It extends C-Pilot to look at the maintainability and reliability impact of the code that's changed right inside the pull request itself. We're working with a set of customers to make sure we're providing something useful for engineering leads and something that's actually loved by developers. So, the team's excited to get this into public preview, hear your feedback, and help keep your code high quality at scale. Next up, something I know I've gotten plenty of feedback from our customers on is co-pilot metrics. Today, I'm pleased to announce the public preview of C-Pilot metrics dashboard. This dashboard shows the impact of using C-Pilot and any coding agent across your entire organization, helping give you access to the information about how your company is using C-Pilot. This gives you muchneeded guidance and advice on what makes sense to measure to drive real impact as well as an easy way to get started with a dashboard and dive deep with API access. You can use the C-pilot metrics dashboard today with public preview with your existing Copilot business or Copilot Enterprise subscriptions. Finally, we need to make sure you have the AI controls you need for your enterprise to manage your C-pilot deployment and use all of what we shared today, new models and future features as we build them. With AI controls, we're giving you a clear control plane for co-pilot. and in the future across all of Agent HQ. All of this from a single enterprisewide admin interface available to administrators and to trusted delegates via a new custom role. Today, we're starting with highly requested controls to centrally manage your agents and define a shared location for custom agent profiles to be stored. View agent audit logs so you know which user kicked off an agent, what that agent did, and to track it down like you do other events on GitHub and control what parts of your codebase agents can and cannot access and which features are available to your teams. These new AI controls show our renewed commitment to enterprises so they can turn on the features they want with increasingly granular control all in a single place. [Music], GitHub Universe 2025 Day 1 Announcements, YouTube, 01k8rtpw6j7evbddyc3rcnb5w5.