GitHub Agent HQ adds Claude and Codex to let developers run multiple AI coders side by side

GitHub is widening its bet on agent based development. The company has added Claude from Anthropic and OpenAI Codex to Agent HQ, making both available in public preview for users on Copilot Pro+ and Copilot Enterprise across GitHub and Visual Studio Code.

Agent HQ is GitHub’s control layer for AI coding agents. Instead of treating models as separate tools, it brings them into a single workspace where context, history, issues, and pull requests stay tied to the repository. With the latest update, developers can run Copilot, Claude, and Codex in parallel, assign them different parts of a task, or ask them to tackle the same problem and compare how each one reasons through tradeoffs.

Agent HQ keeps AI work inside the repo

The key shift is where the work happens. Agents operate directly on issues and pull requests, producing draft PRs and comments that can be reviewed like any teammate’s contribution. GitHub says this makes it easier to test competing approaches, pressure test logic for edge cases, and propose smaller, safer refactors without losing context to copy paste workflows.

For teams already using AI daily, Agent HQ is shaping up as less chatbot and more command center.