GitHub Agent HQ

Agent HQ is GitHub’s way of treating coding agents as a team you manage instead of a single assistant. Rather than choosing one AI vendor, a developer can assign work to agents from different providers, such as OpenAI’s Codex and Anthropic’s Claude Code, from the same place where code, issues, and pull requests already live. Each agent works on its task and returns changes for review in the usual pull request flow.

Mission Control gives a central view of what every agent is doing, so a developer can see which tasks are running, which are finished, and which need attention. Plan Mode lets an agent propose its steps before writing code, so the developer can correct the approach early. Because it works across GitHub, the command line, and VS Code, the same tasks can be started and tracked from whichever surface fits.

Organizations get controls over which agents and models their teams may use, and a metrics dashboard shows usage and results. Access depends on a paid Copilot plan, and some agents may bill premium requests. Running multiple agents increases the amount of code to review, so review practices matter more, not less. Cursor and Devin offer their own agent management, and Claude Code and Codex can be run directly. Agent HQ appeals most to teams already centered on GitHub.