Agent HQ is GitHub’s name for making coding agents native to the GitHub flow rather than something bolted onto it. Announced at GitHub Universe 2025 (see Introducing Agent HQ: Any agent, any way you work), it is described by GitHub as an open ecosystem that unites agents from several vendors on one platform, reached through a paid GitHub Copilot subscription instead of through each vendor’s own surface. GitHub states that agents from Anthropic, OpenAI, Google, Cognition and xAI would become available this way over the months following the announcement.
What Agent HQ deliberately leaves unchanged is part of how GitHub defines it: the working primitives stay Git, pull requests and issues, and the compute stays GitHub Actions or self-hosted runners. The layer it adds is direction and governance — one place to assign and track agent work, editor-side controls over how an agent behaves, and administrative controls over which agents may run at all.
Capabilities
- Mission control — a command center GitHub describes as a consistent interface across GitHub, VS Code, mobile and the CLI rather than a single destination. From it a developer can pick from a fleet of agents, assign work in parallel, and track progress from any device.
- Branch controls giving granular oversight of when CI and other checks run against agent-created code.
- Agent identity, so that which agent is building a task, its access, and the policies applied to it are managed as they would be for a human team member.
- One-click merge conflict resolution, improved file navigation, and code-commenting improvements.
- Integrations with Slack and Linear, alongside previously announced connections for Atlassian Jira, Microsoft Teams, Azure Boards and Raycast.
- Plan Mode in VS Code, where Copilot asks clarifying questions to build a step-by-step approach before implementation; once the plan is approved it is handed to Copilot to implement, locally or through a cloud agent.
- Custom agents configured through AGENTS.md files — source-controlled documents carrying rules and guardrails that shape an agent’s behaviour without re-prompting — and custom agents in GitHub Copilot with their own system prompt and tools.
- GitHub MCP Registry in VS Code, for discovering, installing and enabling Model Context Protocol servers with a single click. GitHub states that VS Code is the only editor supporting the full MCP specification.
- Agent control plane — the governance layer for enterprise administrators, covering security policies, audit logging, access management, which agents are permitted, and which models they may reach. In public preview at announcement.
- Copilot metrics dashboard, in public preview at announcement, reporting Copilot usage and impact across an organization.
Adoption & Ecosystem
GitHub positions Agent HQ against a problem it states in terms of code quality rather than capability: a review can pass — “LGTM” — while the change still degrades the codebase into long-term technical debt. GitHub Code Quality, announced in public preview alongside Agent HQ, extends Copilot’s security checks to the maintainability and reliability impact of changed code and adds org-wide visibility and reporting. Separately, a code review step was added inside GitHub Copilot coding agent’s own workflow so that its output receives a first-line review before a human sees it.
The first partner agent to arrive was OpenAI Codex, which GitHub made available to Copilot Pro+ users in VS Code Insiders in the week of the announcement — described as the first of the partner agents to extend beyond its own native surfaces into the editor.
All of the above is GitHub’s own account of a platform it was announcing rather than an independent report of shipped behaviour, and several capabilities are stated as forthcoming or in public preview at the time of writing.