With more than 180 million developers, a new developer joining the site every single second, 80% of new sign-ups using Copilot in their first week, and AI now “an integral part of the development experience,” GitHub is betting its future on a world where developers orchestrate fleets of specialised agents rather than juggle one-off chatbots or extensions.

“Agent HQ isn’t about the hype of AI. It’s about the reality of shipping code,” said GitHub COO Kyle Daigle (pictured) on stage at GitHub Universe 2025 in San Francisco. “We built Agent HQ because we’re developers, too ... When ‘AI-powered’ ends up meaning more context-switching, more babysitting, more subscriptions, and more time explaining what you need - that ends today.”

At its heart, Agent HQ reframes GitHub as a platform of platforms: a single, governed environment where multiple autonomous agents can run, collaborate and be held accountable. That includes GitHub’s own Copilot Coding Agent, plus new integrations arriving over the next few months from Anthropic (Claude), OpenAI (Codex), Google, Cognition, and xAI - all available through a developer’s existing paid Copilot subscription.

Rather than bolting agents onto separate products, GitHub has embedded agent orchestration directly into the GitHub Flow - issues, branches, pull requests, reviews - so existing workflows, permissions and audit trails remain intact.

Mission Control: one view across every agent, anywhere

The centrepiece is Mission Control, a unified command interface that follows developers across GitHub .com, VS Code, the CLI and even mobile. From Mission Control, a developer can see every active agent session, assign new tasks, watch progress in real time and review results.

Agents can be given distinct identities, limited permissions and granular branch controls - for instance, when to trigger CI, security scans or dependency checks. GitHub even added one-click merge-conflict resolution and new comment/review tools designed for code generated by agents.

Teams can also manage agent-related work directly from Slack, Microsoft Teams, Linear, Jira, Azure Boards and Raycast, ensuring AI-driven work stays within normal communication and tracking systems.

Inside VS Code: plan, customise, and connect

Microsoft and GitHub have closely collaborated on extending Agent HQ deeply into Visual Studio Code - also announcing VS Code is now officially GitHub’s “flagship editor”. The newest VS Code release brings several powerful workflows:

• Plan Mode. Developers collaborate with Copilot to draft a step-by-step plan for a feature or fix. Copilot asks clarifying questions, finds missing decisions and gaps, then converts the approved plan into executable work - locally or through cloud agents.
• Custom Agents. Teams can create AGENTS.md files - source-controlled configuration documents specifying prompts, tools and rules (e.g. “use table-driven tests for all handlers”). These shape Copilot’s behaviour without repetitive prompting.
• Full MCP support. VS Code now supports the entire Model Context Protocol spec and includes the new GitHub MCP Registry. Developers can one-click install vetted MCP servers such as Stripe, Figma, Sentry or internal tools, giving agents direct, policy-controlled access to specialised APIs.
• OpenAI Codex arrives first. From this week, Copilot Pro Plus users on VS Code Insiders can use Codex natively inside GitHub workflows. Codex is tuned for long-running, parallel tasks - engineers have reported letting it tackle complex refactors for 60 hours straight.

“Together with GitHub, we’re meeting developers wherever they work,” said OpenAI Codex product lead Alexander Embiricos. “The first Codex model helped power Copilot and inspired a new generation of AI-assisted coding. Now we’re extending that power everywhere code gets written.”

Anthropic chief product officer (and also Instagram co-founder) Mike Krieger echoed the sentiment: “With Agent HQ, Claude can pick up issues, create branches, commit code, and respond to pull requests - working alongside your team like any other collaborator.”

Confidence, control and code quality

GitHub knows enterprise adoption depends on more than speed. Three new governance pillars round out Agent HQ:

1. Code Quality (public preview). Expands Copilot’s checks to measure maintainability, reliability and test coverage impact per PR - not just whether code “looks good.” Provides org-wide dashboards for engineering leads.
2. Copilot Metrics Dashboard. Gives visibility into usage, adoption and ROI - critical for IT leaders justifying AI spend to CFOs and boards.
3. AI Control Plane. A unified console where administrators can set security policies, limit model/agent access, enable audit logging, and manage custom agent profiles.

Agent HQ’s self-hosted runner support means even regulated enterprises can run agent workloads within their own Azure, AWS or on-prem environments, satisfying governance frameworks such as APRA CPS 234, ISO 27001, or Australia’s ASD Essential Eight.

Why it matters to Australian developers and CIOs

Australia’s public sector and finance industries have been accelerating adoption of AI-assisted software delivery but remain cautious about governance. The Digital Transformation Agency’s AI Technical Standard and APRA’s security standards both stress accountability, transparency and auditability - exactly what Agent HQ’s control plane and per-agent identity model address.

Local GitHub enterprise customers such as Atlassian, Canva, and major banks can now let developers experiment with Copilot and third-party agents while keeping telemetry, code review, and data residency under local control.

Independent Australian developers also benefit: VS Code Insiders builds are already pushing weekly updates with Plan Mode and MCP Registry, giving startups and open-source maintainers access to the same orchestration stack as global enterprises.

GitHub VP APAC, India, China and Japan Sharryn Napier told iTWire GitHub now has more than two million Australian developers on the platform. And, in fact, Australia's own Commonwealth Bank was the first organisation in the world to use GitHub's coding agent.

The bigger picture: orchestration, not autocomplete

Agent HQ signals a philosophical shift: from “AI writes code for you” to “AI works with you.” Developers aren’t expected to surrender control but to act as conductors of an intelligent, auditable orchestra.

By anchoring agents inside existing GitHub primitives - Git, Issues, Pull Requests, Actions - GitHub keeps trust and traceability intact while absorbing the best ideas from emerging rivals. It’s a Switzerland of software agents, where OpenAI, Anthropic and even Google plug in rather than compete for editor real estate.

For GitHub, that means insulation from the volatility of model wars; for developers, it means fewer tabs, fewer integrations, and finally a chance for AI to reduce complexity instead of multiplying it.

GitHub isn't chasing hype; it's systematising it. Agent HQ brings order, auditability, and choice to AI-assisted development. It is the start of a genuinely new era: the age of agentic software engineering, governed and grounded on GitHub.

"It's about giving the power to build faster, with more confidence, and on your terms," Daigle said.