Claude and Codex Now Available as Coding Agents on GitHub (Agent HQ)
GitHub CopilotGitHub has opened its platform to third-party AI coding agents, making Anthropic's Claude and OpenAI's Codex available in public preview directly on github.com for Copilot Pro+ and Copilot Enterprise subscribers. Developers can now assign tasks to Claude, Codex, or GitHub Copilot — or all three simultaneously — from issues, pull requests, and a dedicated Agents tab without leaving the GitHub interface. Each coding agent session consumes one premium request during the public preview period, with no additional subscription required beyond an existing Copilot plan.
Sources & Mentions
5 external resources covering this update
GitHub previews support for Claude and Codex coding agents
InfoWorld
GitHub is letting developers choose between Copilot and its biggest rivals
The New Stack
GitHub integrates Claude and Codex AI coding agents directly into GitHub
TechRadar
GitHub Adds Claude Code and OpenAI Codex to Agent HQ Platform
Winbuzzer
Claude and Codex are now available in public preview on GitHub (Community Discussion)
GitHub
GitHub Opens Its Platform to Third-Party Coding Agents
In a significant strategic shift, GitHub has integrated Anthropic's Claude and OpenAI's Codex directly into github.com as first-class coding agents, available in public preview for Copilot Pro+ and Copilot Enterprise subscribers. The feature — accessible through what GitHub calls Agent HQ — allows developers to delegate coding tasks to multiple AI agents from a single, unified interface without context-switching between tools.
How Agent HQ Works
Developers can assign an issue or task to Claude, Codex, GitHub Copilot, or any combination of all three, using the existing Assignees dropdown in GitHub's issue and pull request views. Once assigned, the selected agent automatically begins working and submits a draft pull request for review. Agents are accessible from several surfaces:
- Issues and pull requests on github.com
- The dedicated Agents tab in enabled repositories
- The agent sessions view in VS Code (version 1.109 or later required)
- GitHub Mobile on iOS and Android
Multi-Agent Comparison and Iteration
One of the most notable capabilities of Agent HQ is the ability to run multiple agents on the same task simultaneously. Developers can assign a single issue to Claude, Codex, and Copilot in parallel and compare how each agent reasons through the problem, weighing trade-offs and arriving at different implementation strategies. Agents support iterative refinement — developers can leave feedback on a draft pull request and the agent will revise its approach across multiple rounds.
Availability and Pricing
Access to Claude and Codex is included with existing Copilot Pro+ and Copilot Enterprise subscriptions — no additional licensing is required. During the public preview period, each coding agent session consumes one premium request. GitHub notes that enablement steps differ by plan: Pro+ users configure agent access per repository under Copilot coding agent settings and toggle Claude and Codex individually, while Enterprise users require administrators to enable agents at both the enterprise and organization levels under AI Controls and Copilot settings. VS Code users must be on version 1.109 or later to access agent sessions.
GitHub confirmed that additional third-party agents from Google, Cognition, and xAI are in the pipeline for future releases.