GitHub Copilot больше не монополист: Claude теперь официально в деле

Microsoft and GitHub have done something that seemed unthinkable a year ago for a corporate giant: they voluntarily dismantled the monopoly of their own model in the main tool of the modern programmer. GitHub Copilot ceased to be merely a wrapper around GPT and became a full-fledged platform. Now developers can choose the "brains" for their AI assistant as easily as selecting a theme in a code editor. The list of available engines officially includes Claude from Anthropic and Codex from OpenAI, marking the end of an era of exclusivity.

To understand why this is happening now, you need to look at the market. Over the past six months, the industry has been shaken by the success of Claude 3.5 Sonnet. This model from Anthropic turned out to be so good at logic and writing clean code that even the most devoted OpenAI fans began switching to alternative IDEs like Cursor. GitHub, being the historical leader, found itself in the position of a follower. Developers no longer want to be locked into the confines of one, however powerful, neural network. They need flexibility and the best tool for the specific task.

The new update, dubbed Agent HQ, is GitHub's attempt to become not just a "smart notepad," but a mission control center for an entire army of AI agents. Now in Visual Studio Code, mobile app, and GitHub's web interface, there is the ability to switch between models on the fly. This is part of a grand vision of the future where AI agents become native participants in the development process. You no longer ask a chatbot to "write a function," you assign an agent to a specific task in the backlog.

The mechanics of operation have become significantly deeper. Now a developer can attach a specific agent to a ticket or pull request. The agent analyzes the entire codebase, understands the context of dependencies, and proposes a ready-made solution that can be merged into the project immediately. This is a qualitative shift from code-line suggestions to autonomous task execution. Of course, such luxury costs money: access to model selection is only available to subscribers of Copilot Pro Plus and Copilot Enterprise plans. Each agent action consumes so-called premium requests, turning AI usage into a measurable production resource.

For Microsoft, this move looks like a subtle political game. On the one hand, they acknowledge the power of Anthropic—their direct competitor in which Amazon and Google are actively investing. On the other hand, by keeping the developer inside VS Code and GitHub, Microsoft retains control over the infrastructure. It matters less to them whose model generates tokens if all development, hosting, and deployment pipelines happen in their Azure cloud. This is a classic "embrace and extend" strategy, adapted to the realities of the AI boom.

It's important to note that Codex, mentioned in the update, is not merely a return to old roots, but an optimized version for those accustomed to the specific style of OpenAI model responses. However, the main focus is on Claude. The integration of Sonnet 3.5 into the GitHub ecosystem makes this alliance the most powerful tool on the market at the present time. This is a direct challenge to all startups that tried to take market share through multi-model capabilities.

Ultimately, we are seeing GitHub transform into an operating system for development, where neural networks are simply drivers. Today there are two or three of them, tomorrow there will be dozens. Perhaps soon we will see specialized models, trained exclusively on security or optimization of high-load systems. GitHub has created the foundation for this future by abandoning the idea that one model can rule them all.

The bottom line: GitHub stopped fighting reality and recognized Claude as the developers' favorite. Now it is not just a service, but an open marketplace of AI agents, where Microsoft plans to earn money on commissions and infrastructure, not exclusivity.