GitHub adds Claude and Codex to Copilot

According to GitHub's Feb 26, 2026 blog post, GitHub made third-party coding agents Claude (Anthropic) and Codex (OpenAI) available to Copilot Business and Copilot Pro customers, expanding earlier access for Copilot Enterprise and Pro+ users. The announcement states developers can run these agents inside github.com, GitHub Mobile, and Visual Studio Code, with agent sessions sharing history and context across surfaces. GitHub's post adds that access to Claude and Codex is "No additional subscriptions are required," and that each coding agent session consumes one premium request during public preview. The post also notes the Agent Control Plane is generally available for centralized enablement, policy management, and audit logging.

The GitHub post and Visual Studio Code documentation describe how third-party agents integrate into the existing Copilot platform. Per the VS Code docs, cloud-based third-party agents use the provider's SDK and an agent harness to expose each provider's capabilities inside the editor, and billing/authentication is handled through the user's Copilot subscription. The VS Code documentation specifically references a configuration flag github.copilot.chat.claudeAgent.enabled for enabling Claude sessions at organization level. GitHub's announcement lists agent access to repository code and history, issues and pull requests, Copilot Memory, and repository instructions and policies; it also describes agent features such as opening draft pull requests, assigning issues to agents, asynchronous sessions, and detailed logs for review.

Industry-pattern observations: Integrating multiple third-party agents into a single developer UX follows a broader trend of "multi-agent" and "multi-model" experiences where platform providers let users choose between specialized models without changing tools. For practitioners, this reduces friction when comparing outputs from models with different strengths, and it centralizes governance, auditing, and billing under the platform rather than at the individual provider level. Operators and security teams will likely view the Agent Control Plane as a meaningful control point, since GitHub reports centralized policy management and audit logging for agent activity.

Editorial analysis: The move places model choice inside an established developer workflow, rather than forcing teams to adopt separate provider consoles or editor extensions. For teams already standardized on GitHub and VS Code, the integration lowers the switching cost to evaluate Claude or Codex on real project data. Public reporting from Tessl and others frames the change as part of a crowded AI coding landscape where multiple vendors offer agentic coding assistants; the practical implication is that differentiation will increasingly rest on model behavior in repository context, toolchain ergonomics, and enterprise controls rather than simple availability.

For practitioners: Monitor how billing and quota consumption behave as third-party agents move out of preview, and watch which agent types consume premium requests under different Copilot tiers. Observers should also watch audit logs and policy surfaces exposed by the Agent Control Plane to evaluate compliance and data governance.