GitHub has taken a major step forward in AI-assisted software development by adding Anthropic’s Claude and OpenAI’s Codex as first-class coding agents directly inside the platform.

Instead of operating as detached chatbots, these AI agents work exactly where developers already collaborate—inside repositories, issues, and pull requests.

The update transforms GitHub from a place where AI assists developers into a platform where AI can actively do the work, while still following standard review and governance workflows.

What’s New in GitHub’s AI Agent Experience

The new feature, rolling out in public preview, is available to Copilot Pro Plus and Copilot Enterprise subscribers. Access is supported across:

• GitHub’s web interface
• GitHub mobile app
• Visual Studio Code

No additional subscriptions are required, although each agent session counts as a premium Copilot request.

Unlike earlier Copilot integrations—where developers switched between chat windows and code editors—these agents are embedded directly into GitHub’s core workflows.

Agent HQ: Treating AI Like Real Team Members

GitHub calls this new system Agent HQ, and the philosophy is simple: AI agents are treated as discrete workers rather than endless conversational tools.

Developers can assign issues directly to:

• GitHub Copilot
• Anthropic Claude
• OpenAI Codex

Each agent works independently and can submit draft pull requests, complete with code changes and explanations. Multiple agents can even be assigned to the same issue, allowing teams to compare solutions side by side before choosing the best approach.

Every action stays tied to the repository, creating a transparent trail of decisions, discussions, and changes.

Why GitHub Is Embracing Multiple AI Agents

Different AI models excel at different tasks. Some are better at refactoring legacy code, while others perform best when building new features from scratch. GitHub’s multi-agent approach allows developers to choose the right AI for each job rather than committing to a single provider.

Mario Rodriguez, GitHub’s Chief Product Officer, explains that the goal is reducing context switching. Moving between IDEs, chat tools, and repositories disrupts focus. Keeping everything in one place preserves momentum and keeps work history, conversations, and code tightly connected.

Despite Microsoft’s heavy investment in OpenAI, opening GitHub to competing agents signals a bigger priority: positioning GitHub as the central hub for modern software development—regardless of which AI powers the work.

How GitHub AI Agents Work in Practice

AI agents operate asynchronously by default. A developer assigns a task, and the agent works in the background. Progress can be monitored live or reviewed later once the task is complete.

The outputs are tangible and reviewable, including:

• Proposed code changes
• Pull request comments
• Suggested fixes or optimizations

Agents can also be invoked directly within existing pull requests. Mentioning @claude or @codex in a review comment triggers analysis and feedback, all of which remains visible to the entire team.

Using Claude, Codex, and Copilot Together

Each agent brings distinct strengths:

• Claude focuses on structured reasoning and iterative improvement, often explaining why changes are proposed rather than simply generating code.
• Codex returns to its roots as a code-focused model, allowing developers to directly compare its output against Copilot’s newer capabilities.
• Copilot continues to evolve as GitHub’s default AI assistant, now competing head-to-head with other agents in the same workspace.

This side-by-side comparison helps teams make decisions based on real results instead of vendor promises.

What This Means for Development Teams

The biggest immediate benefit is choice and transparency. Teams can evaluate how different AI agents approach the same problem and adopt the one that consistently delivers the best results for specific tasks.

From a governance perspective, keeping AI-generated code inside GitHub solves a major problem. When code is created in external tools, tracking its origin and intent becomes difficult. Agent HQ ensures AI contributions follow the same review standards as human work.

GitHub plans to expand access beyond current subscription tiers and is already collaborating with Google, Cognition, and xAI to bring more agents onto the platform.

The Bigger Picture for GitHub and AI Development

Microsoft has reportedly been testing Anthropic’s Claude internally, comparing it directly with Copilot to gather insights that help improve GitHub’s own tools. Even as competition increases, those lessons feed back into the ecosystem.

GitHub’s direction is clear: it wants to be the place where AI agents actually build software, not just talk about it. As AI-driven development accelerates, the platform is positioning itself at the center of that future.