Back

GitHub Introduces Claude and Codex AI Coding Agents in Public Preview

GitHub Introduces Claude and Codex AI Coding Agents in Public Preview
The Verge

New AI Coding Agents Integrated into GitHub

GitHub has announced a public preview that makes two rival AI coding agents—Claude from Anthropic and Codex from OpenAI—directly accessible inside the core GitHub experience, as well as in GitHub Mobile and Visual Studio Code. The preview is currently limited to users who hold a Copilot Pro Plus or Copilot Enterprise subscription.

Agent HQ and the Choice of Multiple Models

The rollout is part of GitHub’s broader "Agent HQ" strategy, which seeks to embed AI agents into the everyday workflow of developers. Within this framework, developers can now choose between Copilot, Claude, Codex, or any custom‑built agents when creating or managing tasks. Each interaction with an agent consumes a premium request, and users can assign agents to specific issues and pull requests.

Performance Comparison and Context Preservation

GitHub is giving developers tools to evaluate how each AI agent performs on a given problem. By allowing side‑by‑side comparisons, the platform helps developers decide which model best fits their needs for a particular coding step. Mario Rodriguez, GitHub’s chief product officer, emphasized that reducing context switching is essential for smoother software development, noting that the ability to move from idea to implementation without leaving the GitHub environment lessens friction.

Expanding the Ecosystem of AI Models

GitHub has already integrated models from several leading AI providers, including Anthropic, Google, xAI, and OpenAI, into its Copilot offering. The addition of Claude and Codex expands this ecosystem, and GitHub plans to extend access to these agents to more Copilot subscription tiers in the near future. Ongoing collaborations with Google, Cognition, and xAI aim to bring additional agents to GitHub, Visual Studio Code, and the Copilot CLI.

Microsoft’s Parallel Evaluation

Microsoft, GitHub’s parent company, is simultaneously testing Anthropic’s Claude Code tool. Internal developers have been asked to compare Claude Code with GitHub Copilot, a move intended to inform future enhancements to Copilot’s capabilities.

Looking Ahead

The integration of Claude and Codex marks a significant step toward a more flexible, model‑agnostic AI environment for developers. By offering multiple agents within a single platform, GitHub hopes to empower developers to select the most effective tool for each coding challenge, streamline workflows, and ultimately accelerate software delivery.

Used: News Factory APP - news discovery and automation - ChatGPT for Business

Source: The Verge

Also available in: