GitHub Copilot Agent HQ, Explained: One Pen, Every Agent
Every lab shipped a coding agent. GitHub shipped the pen to herd them all. Here's what Agent HQ, mission control, and the new Copilot app actually are — and who gets access.
- GitHub announced Agent HQ on October 28, 2025: an open ecosystem uniting coding agents from Anthropic, OpenAI, Google, Cognition, and xAI, included with paid GitHub Copilot subscriptions.
- Anthropic's Claude and OpenAI's Codex agents hit public preview on Agent HQ on February 4, 2026 for Copilot Pro+ and Enterprise — you can assign both to one task and compare their solutions.
- The GitHub Copilot desktop app (technical preview, June 2, 2026) runs every session in its own isolated git worktree; the Copilot cloud agent runs on all paid plans in ephemeral GitHub Actions environments.
- Since June 1, 2026, agent usage burns GitHub AI Credits per token at published API rates: Copilot Pro is $10/month with $10 in credits, Pro+ is $39 with $39.
GitHub Copilot Agent HQ, announced October 28, 2025, is GitHub's open ecosystem for running coding agents from Anthropic, OpenAI, Google, Cognition, and xAI inside paid Copilot subscriptions. Its mission control interface assigns, monitors, and steers agent tasks across GitHub, VS Code, mobile, and the CLI; Anthropic's Claude and OpenAI's Codex entered public preview February 4, 2026.
Every lab shipped a coding agent this year. GitHub shipped the pen.
GitHub Copilot Agent HQ is GitHub’s open ecosystem for running coding agents from Anthropic, OpenAI, Google, Cognition, and xAI on one platform, included with paid GitHub Copilot subscriptions. GitHub announced it on October 28, 2025, along with mission control — the interface that lets you “direct, monitor, and manage every AI-driven task” across GitHub, VS Code, mobile, and the CLI.
I keep telling y’all… it’s a developer’s responsibility to be able to herd their agents. GitHub heard that and built the herding pen where the pull requests already live.
What is GitHub Copilot Agent HQ, exactly?
Agent HQ is the ecosystem — the platform layer that puts third-party coding agents inside GitHub Copilot. Mission control is its command center, with a web entry at github.com/copilot/agents. The GitHub Copilot app is the desktop layer, in technical preview since June 2, 2026. The Copilot cloud agent is the execution engine that actually does the work.
Four names, four layers. Learn the stack once and the confusion dies.
The same announcement shipped VS Code Plan Mode, custom agents via AGENTS.md, and the GitHub MCP Registry. But peep the posture: GitHub isn’t betting on one model winning. They’re selling the stadium.
Can you run Claude and Codex agents on GitHub Agent HQ?
Yes. Anthropic’s Claude agent and OpenAI’s Codex agent went into public preview on Agent HQ on February 4, 2026 for Copilot Pro+ and Copilot Enterprise subscribers, on github.com, GitHub Mobile, and VS Code 1.109+. You invoke them from the Agents tab, assign them to issues, or mention @Claude or @Codex in a pull request comment.
The detail that got me: assign multiple agents to one task and watch Copilot, Claude, and Codex “reason about tradeoffs and arrive at different solutions,” per GitHub’s own post. That’s the assembly line I keep preaching… One model to dream. One model to build. One model to audit.
Since April 14, 2026 there’s real model routing, per GitHub’s changelog — Claude Sonnet 4.6 or Opus 4.6 on one side, GPT-5.2-Codex up through GPT-5.4 on the other. Google, Cognition, and xAI? Still listed as partners GitHub is “actively working with.” Two agents live, three on the guest list.
What’s the difference between the Copilot app and the cloud agent?
The GitHub Copilot app is the desktop layer — an agent-native app in technical preview since June 2, 2026 for Copilot Pro, Pro+, Business, and Enterprise, where every session runs in its own git worktree. The Copilot cloud agent is available on all paid plans and runs each task in an ephemeral environment powered by GitHub Actions, then opens pull requests.
I got my start in Ops, and this design reads like somebody who’s been paged at 3 a.m. wrote the spec: local worktrees so agents can’t stomp each other’s branches, fully isolated ephemeral Linux sandboxes in the cloud, and Agent Merge babysitting CI — monitoring checks, tracking reviewers, addressing failures before you look up.
That’s blast-radius thinking. That’s the difference between a demo and a platform.
Should you use Agent HQ instead of Claude Code or Cursor?
If you already pay for Copilot Pro+ or Enterprise, Agent HQ is the cheapest way to run Claude and Codex side by side — no new subscriptions, no context-switching out of your pull requests. If your workflow lives inside one tool, like Claude Code in the terminal or Cursor in the editor, Agent HQ complements it. It doesn’t replace it yet.
Full transparency: my daily drivers are Claude Code and Codex, not GitHub Copilot, so read this as informed comparison shopping — I just do my herding in the terminal. If the category is still fuzzy, start with what an AI agent actually is, then peep my best AI agents of 2026 rankings for the full roster.
Money talk. Since June 1, 2026, Copilot bills through GitHub AI Credits — per token, at published API rates. Pro is $10/month with $10 in credits, Pro+ is $39 with $39. Completions stay free; agent sessions eat. Watch your burn.
And clock the timing: SpaceX agreed to acquire Cursor for $60 billion on June 16 — two weeks after the Copilot app dropped, eight months after Agent HQ. GitHub didn’t build this as a reaction; the dates don’t allow it. But the fight for where developers herd their agents is on.
Agents got cheap. Herding them didn’t… GitHub figured out the pen is the product.
#TheAIMogul
Bottom lineAgent HQ turns GitHub into the herding pen for coding agents — Claude and Codex side by side inside one Copilot subscription. If your work lives in GitHub pull requests, it's the most ops-serious orchestration play shipping today. It doesn't replace Claude Code or Cursor yet.