Introducing Agent HQ: Any agent, any way you work

The existent AI scenery presents a situation we’re each excessively acquainted with: unthinkable powerfulness fragmented crossed antithetic tools and interfaces. At GitHub, we’ve ever worked to lick these kinds of systemic challenges—by making Git accessible, codification reappraisal systematic with propulsion requests, and automating deployment with Actions. 

With 180 cardinal developers, GitHub is increasing astatine its fastest complaint ever—a caller developer joining each second. What’s more, 80% of caller developers are utilizing Copilot successful their archetypal week. AI isn’t conscionable a instrumentality anymore; it’s an integral portion of the improvement experience. Our work is to guarantee this caller epoch of collaboration is powerful, secure, and seamlessly integrated into the workflow you already trust.

At GitHub Universe, we’re announcing Agent HQ, GitHub’s imaginativeness for the adjacent improvement of our platform. Agents shouldn’t beryllium bolted on. They should enactment the mode you already work. That’s wherefore we’re making agents autochthonal to the GitHub flow.

Agent HQ transforms GitHub into an unfastened ecosystem that unites each cause connected a azygous platform. Over the coming months, coding agents from Anthropic, OpenAI, Google, Cognition, xAI, and much volition go disposable straight wrong GitHub arsenic portion of your paid GitHub Copilot subscription.

To bring this imaginativeness to life, we’re shipping a suite of caller capabilities built connected the primitives you trust. This starts with a mission control, a azygous bid halfway to assign, steer, and way the enactment of aggregate agents from anywhere. It extends to VS Code with caller ways to program and customize cause behavior. And it is backed by enterprise-grade functionality: a caller procreation of agentic codification review, a dedicated control level to govern AI entree and cause behavior, and a metrics dashboard to recognize the interaction of AI connected your work. 

We are besides profoundly committed to investing successful our level and strengthening the primitives you trust connected each day. This caller satellite of improvement is powered by that foundational work, and we look guardant to sharing much updates. 

Let’s dive in.

GitHub is your Agent HQ: An unfastened ecosystem for each agents  

The aboriginal is astir giving you the powerfulness to orchestrate a fleet of specialized agents to execute analyzable tasks successful parallel, not juggling a patchwork of disconnected tools oregon relying connected a azygous agent. As the pioneer of asynchronous collaboration, we judge it’s our work to marque definite these next-generation async tools conscionable work. 

With Agent HQ what’s not changing is conscionable arsenic important arsenic what is. You’re inactive moving with the primitives you know—Git, propulsion requests, issues—and utilizing your preferred compute, whether that’s GitHub Actions oregon self-hosted runners. You’re accessing agents done your existing paid Copilot subscription. 

On apical of that foundation, we’re opening the doors to a caller satellite of capability. Over the coming months, coding agents from Anthropic, OpenAI, Google, Cognition, and xAI volition beryllium disposable connected GitHub as portion of your paid GitHub Copilot subscription.

Don’t privation to wait? Starting this week, Copilot Pro+ users tin statesman moving with  OpenAI Codex successful VS Code Insiders, the archetypal of our spouse agents to widen beyond its autochthonal surfaces and straight into the editor.

Mission control: Your bid center, wherever you build

The powerfulness of Agent HQ comes from mission control, a unified bid halfway that follows you wherever you work. It’s not a azygous destination; it’s a accordant interface crossed GitHub, VS Code, mobile, and the CLI that lets you direct, monitor, and negociate each AI-driven task. With ngo control, you tin take from a fleet of agents, delegate them enactment successful parallel, and way their advancement from immoderate device. 

We’re besides providing: 

• New branch controls that springiness you granular oversight implicit erstwhile to tally CI and different checks for agent-created code.
• Identity features to power which cause is gathering the task, managing access, and policies conscionable similar you would with immoderate different developer connected your team.
• One-click merge struggle resolution, improved record navigation, and amended codification commenting capabilities.
• New integrations for Slack and Linear, connected apical of our precocious announced connections for Atlassian Jira, Microsoft Teams and Azure Boards, and Raycast. 

Try ngo power today.

New successful VS Code: Plan, customize, and connect

Mission power is successful VS Code, too, truthful you’ve got a azygous presumption of each your agents moving successful VS Code, successful the Copilot CLI, oregon connected GitHub.

Today’s brand caller merchandise successful VS Code is each astir moving alongside agents connected projects, and it’s not astonishing that large results commencement with a large plan. Getting the discourse close earlier a task is critical, but that aforesaid discourse needs to transportation done into the work. Copilot already adapts to the mode your squad works by learning from your files and your project’s culture, but sometimes you request much pointed context.

So today, we’re introducing Plan Mode, which works with Copilot, and asks you clarifying questions on the way, to assistance you to physique a step-by-step attack for your task. Providing the discourse upfront improves what Copilot tin bash and helps you find gaps, missing decisions, oregon task deficiencies aboriginal successful the process—before immoderate codification is written. Once you approve, your program goes to Copilot to commencement implementing, whether that’s locally successful VS Code oregon utilizing an cause successful the cloud.

For adjacent finer control, you tin present make customized agents successful VS Code with AGENTS.md files, source-controlled documents that fto you acceptable wide rules and guardrails specified arsenic “prefer this logger” oregon “use table-driven tests for each handlers.” This shapes Copilot’s behaviour without you re-prompting it each time.

Now you tin trust connected the caller GitHub MCP Registry, disposable straight successful VS Code. VS Code is the only exertion that supports the afloat MCP specification. Discover, install, and alteration MCP servers similar Stripe, Figma, Sentry, and others, with a azygous click. When your task calls for a specialist, make customized agents successful GitHub Copilot with their ain strategy punctual and tools to assistance you specify the ways you privation Copilot to work. 

Increased assurance and power for your team

Agent HQ doesn’t conscionable springiness you much power—it gives you confidence. Ensuring codification quality, knowing AI’s power connected your workflow, and maintaining power implicit however AI interacts with your codebase and enactment are indispensable for your team’s success, and we’re tackling these challenges head-on.

When it comes to codification quality, the halfway occupation is that “LGTM” doesn’t ever mean “the codification is healthy.” A reappraisal tin pass, but tin inactive degrade the codebase and rapidly go semipermanent method debt. With GitHub Code Quality, successful nationalist preview today, you’ve got org-wide visibility, governance, and reporting to systematically amended codification maintainability, reliability, and trial sum crossed each repository. Enabling it extends Copilot’s information checks to look astatine the maintainability and reliability interaction of the codification that’s been changed.

And we’ve added a codification reappraisal step into the Copilot coding agent’s workflow, too, truthful Copilot gets an archetypal first-line reappraisal and addresses problems (before you adjacent spot the code). 

As an organization, you request to cognize however Copilot is being used. So today, we’re announcing the nationalist preview of the Copilot metrics dashboard, showing Copilot’s interaction and captious usage metrics crossed your full organization.

For endeavor administrators who are managing AI access, including AI agents and MCP, we’re focused connected providing accordant AI controls for teams with the control plane—your cause governance layer. Set information policies, audit logging, and negociate entree each successful 1 place. Enterprise admins tin besides power which agents are allowed, specify entree to models, and get metrics astir the Copilot usage successful your organization.

For developers, by developers 

We built Agent HQ due to the fact that we’re developers, too. We cognize what it’s similar erstwhile it feels similar your tools are fighting you alternatively of helping you. When “AI-powered” ends up meaning much context-switching, much babysitting, much subscriptions, and much clip explaining what you request to get the worth you were promised. 

That ends today.

Agent HQ isn’t astir the hype of AI. It’s astir the world of shipping code.  It’s astir bringing bid and governance to this caller epoch without compromising choice. It’s astir giving you the powerfulness to physique faster, with much confidence, and connected your terms.

Welcome home. Let’s build. 

---