GitHub’s Agent HQ brings agents into the flow of development
At GitHub’s Octoverse event, the company unveiled Agent HQ, a platform-level effort to make coding agents native to the GitHub workflow. The announcement frames this as an open ecosystem where multiple third-party agents can be orchestrated from a single control surface, rather than stitched together across separate tools. GitHub cited its scale—180 million developers, with a new developer joining every second—and noted that 80% of new developers use Copilot in their first week, positioning Agent HQ as an evolution of AI integration across the developer experience. See Octoverse coverage here.
An open ecosystem for partner agents
Agent HQ aims to let organizations run a fleet of specialized agents in parallel. Over the coming months, coding agents from Anthropic, OpenAI, Google, Cognition, and xAI will be made available on GitHub as part of paid Copilot subscriptions. In the nearer term, Copilot Pro+ users can start using OpenAI Codex in VS Code Insiders starting this week; the Insiders build is reachable via the VS Code Insiders link.
Mission control: a consistent command center
Central to Agent HQ is mission control, described as a unified command center that follows work across GitHub, VS Code, mobile, and the CLI. Mission control is designed to:
- assign and steer tasks across multiple agents,
- monitor progress in parallel,
- provide a consistent interface regardless of surface.
Additional controls include branch controls for governing when CI and checks run on agent-created code, identity features to manage which agent performs tasks, and one-click merge conflict resolution. Integrations for collaboration tools such as Slack and Linear join existing connections to Jira, Microsoft Teams, Azure Boards, and Raycast. The mission control changelog is available here.
VS Code: planning, customization, and MCP support
Agent HQ extends into VS Code with several editor-focused additions:
- Plan Mode, which guides agents through clarifying questions to create a step-by-step plan before code is written, then hands the approved plan to Copilot for implementation.
- Support for AGENTS.md files, source-controlled configuration that sets guardrails and preferences (for example, favored logging or test styles) so agent behavior persists across sessions.
- The GitHub MCP Registry available directly in VS Code; VS Code is noted as the only editor supporting the full MCP specification, enabling discovery and installation of MCP servers such as Stripe, Figma, and Sentry. Links: AGENTS.md docs, MCP Registry docs.
Governance, quality, and metrics
Agent HQ pairs new capabilities with governance and observability:
- GitHub Code Quality enters public preview to provide org-wide visibility and to extend Copilot’s checks to include maintainability and reliability impact.
- A new agentic code review step is added to the Copilot coding agent workflow so Copilot performs an initial first-line review of its own outputs.
- The Copilot metrics dashboard is available in public preview to surface Copilot’s impact and usage across organizations.
- Enterprise admins gain a dedicated control plane (in public preview) for policy, audit logging, agent allowance, and model access.
Links: GitHub Code Quality changelog, Copilot metrics dashboard changelog, Enterprise AI controls (agent control plane) changelog.
A pragmatic integration of agents
The Agent HQ announcement emphasizes integrating agents without disrupting established primitives—Git, pull requests, issues, and existing compute such as Actions or self-hosted runners remain central. The effort bundles orchestration, editor-level planning and customization, and enterprise-grade controls to make agents a managed part of the software delivery lifecycle rather than an add-on.
Read the original announcement on the GitHub Blog: https://github.blog/news-insights/company-news/welcome-home-agents
