Users of Copilot Pro+ and Copilot Enterprise can now utilize multiple coding agents seamlessly within GitHub, GitHub Mobile, and Visual Studio Code.
Through its Agent HQ AI platform, GitHub is introducing support for Anthropic Claude and OpenAI Codex coding agents. This functionality is currently available in public preview.
As announced by GitHub on February 4, Copilot Pro+ and Copilot Enterprise subscribers can now execute multiple coding agents directly within GitHub, GitHub Mobile, and Visual Studio Code. Support for Copilot CLI is also expected in the near future, GitHub noted.
The company stated that Agent HQ, integrating Claude, Codex, and GitHub Copilot, empowers developers to transition from concept to execution by leveraging various agents for distinct phases, all without the need to change tools or sacrifice context. Katelyn Lesse, Anthropic’s head of platform, commented, “We are integrating Claude into GitHub to engage developers directly in their workspace. Via Agent HQ, Claude can perform code commits and comment on pull requests, thereby helping teams to accelerate their iteration cycles and deploy with greater assurance. Our objective is to provide developers with essential reasoning capabilities precisely where they are most needed.”
According to GitHub, direct interaction with these agents within GitHub and VS Code enables users to accomplish the following:
- Gain early insights into potential trade-offs by deploying agents concurrently to identify diverse strategies and unusual scenarios before the code becomes finalized.
- Maintain continuous context linked to the project, as agents function directly within the user’s repository, issues, and pull requests, rather than relying on isolated, stateless prompts.
- Streamline existing review workflows; agent-produced modifications appear as draft pull requests and comments, undergoing the identical review process as contributions from a fellow team member.
Furthermore, GitHub mentioned that its Agent HQ allows users to assess how various agents tackle an identical challenge. Developers have the option to allocate several agents to a single task, observing how Copilot, Claude, and Codex analyze potential compromises and formulate diverse solutions.
