The present AI panorama presents a problem we’re all too accustomed to: unbelievable energy fragmented throughout totally different instruments and interfaces. At GitHub, we’ve at all times labored to resolve these sorts of systemic challenges—by making Git accessible, code overview systematic with pull requests, and automating deployment with Actions.
With 180 million builders, GitHub is rising at its quickest price ever—a brand new developer becoming a member of each second. What’s extra, 80% of recent builders are utilizing Copilot of their first week. AI isn’t only a instrument anymore; it’s an integral a part of the event expertise. Our duty is to make sure this new period of collaboration is highly effective, safe, and seamlessly built-in into the workflow you already belief.
At GitHub Universe, we’re asserting Agent HQ, GitHub’s imaginative and prescient for the following evolution of our platform. Brokers shouldn’t be bolted on. They need to work the way in which you already work. That’s why we’re making brokers native to the GitHub circulate.
Agent HQ transforms GitHub into an open ecosystem that unites each agent on a single platform. Over the approaching months, coding brokers from Anthropic, OpenAI, Google, Cognition, xAI, and extra will turn into obtainable straight inside GitHub as a part of your paid GitHub Copilot subscription.
To convey this imaginative and prescient to life, we’re delivery a set of recent capabilities constructed on the primitives you belief. This begins with a mission management, a single command middle to assign, steer, and observe the work of a number of brokers from anyplace. It extends to VS Code with new methods to plan and customise agent conduct. And it’s backed by enterprise-grade performance: a brand new technology of agentic code overview, a devoted management aircraft to manipulate AI entry and agent conduct, and a metrics dashboard to grasp the influence of AI in your work.
We’re additionally deeply dedicated to investing in our platform and strengthening the primitives you depend on day by day. This new world of growth is powered by that foundational work, and we stay up for sharing extra updates.
Let’s dive in.
GitHub is your Agent HQ: An open ecosystem for all brokers
The long run is about supplying you with the facility to orchestrate a fleet of specialised brokers to carry out complicated duties in parallel, not juggling a patchwork of disconnected instruments or counting on a single agent. Because the pioneer of asynchronous collaboration, we imagine it’s our duty to verify these next-generation async instruments simply work.
With Agent HQ what’s not altering is simply as necessary as what is. You’re nonetheless working with the primitives —Git, pull requests, points—and utilizing your most popular compute, whether or not that’s GitHub Actions or self-hosted runners. You’re accessing brokers by means of your current paid Copilot subscription.
On high of that basis, we’re opening the doorways to a brand new world of functionality. Over the approaching months, coding brokers from Anthropic, OpenAI, Google, Cognition, and xAI shall be obtainable on GitHub as a part of your paid GitHub Copilot subscription.
Don’t need to wait? Beginning this week, Copilot Professional+ customers can start working with OpenAI Codex in VS Code Insiders, the primary of our companion brokers to increase past its native surfaces and straight into the editor.
Mission management: Your command middle, wherever you construct
The ability of Agent HQ comes from mission management, a unified command middle that follows you wherever you’re employed. It’s not a single vacation spot; it’s a constant interface throughout GitHub, VS Code, cellular, and the CLI that permits you to direct, monitor, and handle each AI-driven job. With mission management, you possibly can select from a fleet of brokers, assign them work in parallel, and observe their progress from any machine.
We’re additionally offering:
- New department controls that provide you with granular oversight over when to run CI and different checks for agent-created code.
- Identification options to manage which agent is constructing the duty, managing entry, and insurance policies identical to you’d with some other developer in your crew.
- One-click merge battle decision, improved file navigation, and higher code commenting capabilities.
- New integrations for Slack and Linear, on high of our lately introduced connections for Atlassian Jira, Microsoft Groups and Azure Boards, and Raycast.
Attempt mission management at present.
New in VS Code: Plan, customise, and join
Mission management is in VS Code, too, so that you’ve received a single view of all of your brokers working in VS Code, within the Copilot CLI, or on GitHub.
At the moment’s model new launch in VS Code is all about working alongside brokers on tasks, and it’s not shocking that nice outcomes begin with an important plan. Getting the context proper earlier than a mission is essential, however that very same context wants to hold by means of into the work. Copilot already adapts to the way in which your crew works by studying out of your recordsdata and your mission’s tradition, however generally you want extra pointed context.
So at present, we’re introducing Plan Mode, which works with Copilot, and asks you clarifying questions alongside the way in which, that can assist you to construct a step-by-step method to your job. Offering the context upfront improves what Copilot can do and helps you discover gaps, lacking selections, or mission deficiencies early within the course of—earlier than any code is written. When you approve, your plan goes to Copilot to start out implementing, whether or not that’s domestically in VS Code or utilizing an agent within the cloud.
For even finer management, now you can create customized brokers in VS Code with AGENTS.md recordsdata, source-controlled paperwork that allow you to set clear guidelines and guardrails reminiscent of “favor this logger” or “use table-driven assessments for all handlers.” This shapes Copilot’s conduct with out you re-prompting it each time.
Now you possibly can depend on the brand new GitHub MCP Registry, obtainable straight in VS Code. VS Code is the solely editor that helps the total MCP specification. Uncover, set up, and allow MCP servers like Stripe, Figma, Sentry, and others, with a single click on. When your job requires a specialist, create customized brokers in GitHub Copilot with their very own system immediate and instruments that can assist you outline the methods you need Copilot to work.
Elevated confidence and management to your crew
Agent HQ doesn’t simply provide you with extra energy—it offers you confidence. Making certain code high quality, understanding AI’s affect in your workflow, and sustaining management over how AI interacts along with your codebase and group are important to your crew’s success, and we’re tackling these challenges head-on.
With regards to code high quality, the core drawback is that “LGTM” doesn’t at all times imply “the code is wholesome.” A overview can go, however can nonetheless degrade the codebase and shortly turn into long-term technical debt. With GitHub Code High quality, in public preview at present, you’ve received org-wide visibility, governance, and reporting to systematically enhance code maintainability, reliability, and take a look at protection throughout each repository. Enabling it extends Copilot’s safety checks to take a look at the maintainability and reliability influence of the code that’s been modified.
And we’ve added a code overview step into the Copilot coding agent’s workflow, too, so Copilot will get an preliminary first-line overview and addresses issues (earlier than you even see the code).
As a corporation, it’s worthwhile to know the way Copilot is getting used. So at present, we’re asserting the general public preview of the Copilot metrics dashboard, exhibiting Copilot’s influence and demanding utilization metrics throughout your whole group.
For enterprise directors who’re managing AI entry, together with AI brokers and MCP, we’re centered on offering constant AI controls for groups with the management aircraft—your agent governance layer. Set safety insurance policies, audit logging, and handle entry multi function place. Enterprise admins may also management which brokers are allowed, outline entry to fashions, and procure metrics in regards to the Copilot utilization in your group.
For builders, by builders
We constructed Agent HQ as a result of we’re builders, too. We all know what it’s like when it seems like your instruments are combating you rather than serving to you. When “AI-powered” finally ends up which means extra context-switching, extra babysitting, extra subscriptions, and extra time explaining what it’s worthwhile to get the worth you had been promised.
That ends at present.
Agent HQ isn’t in regards to the hype of AI. It’s in regards to the actuality of delivery code. It’s about bringing order and governance to this new period with out compromising alternative. It’s about giving you the facility to construct quicker, with extra confidence, and in your phrases.
Welcome residence. Let’s construct.
Written by
