Microsoft has confirmed plans to cancel most of its Claude Code licenses, the AI coding tool developed by Anthropic, and push its internal developers toward GitHub Copilot CLI instead. The move comes after months of internal testing where Claude Code proved extremely popular among Microsoft engineers — perhaps too popular for the company’s own strategic goals. The decision affects thousands of developers within Microsoft’s Experiences + Devices group, which handles Windows, Office, Teams, and Surface products.
According to internal sources and documents, Microsoft first began offering Claude Code to its developers in December 2025. The tool was introduced to let non-coders — such as project managers and designers — experiment with writing code for the first time. Over the following six months, Claude Code became a central part of many workflows, with employees creating prototypes, generating code snippets, and debugging applications. However, its popularity began to undermine Microsoft’s own AI coding product, GitHub Copilot CLI.
Why Microsoft is stepping away from Claude Code
The transition away from Claude Code is set to be completed by the end of June 2026, which also marks the last day of Microsoft’s fiscal year. Sources indicate that the decision is partly financial: canceling Claude Code licenses is an easy way to cut operating expenses before the new fiscal year starts in July. But the move is also about strategic alignment. An internal memo from Rajesh Jha, executive vice president of Microsoft’s experiences and devices group, stated that while Claude Code was valuable for learning, “Copilot CLI has given us something especially important: a product we can help shape directly with GitHub for Microsoft’s repos, workflows, security expectations, and engineering needs.”
Jha’s memo stressed that Microsoft wants to converge on Copilot CLI as its main agentic command-line interface tool. The company believes it can integrate Copilot CLI more deeply into its own engineering pipelines, security protocols, and development standards. In contrast, Claude Code, while powerful, is a third-party tool that Microsoft cannot customize as freely. By shifting to Copilot CLI, Microsoft gains the ability to prioritize features, fix bugs faster, and align the tool with its own infrastructure.
Internal dynamics and developer sentiment
The transition will not be seamless. Many Microsoft engineers have come to prefer Claude Code over Copilot CLI, citing better performance, more natural language understanding, and a more mature agentic workflow. Over the past half-year, some teams used Claude Code exclusively, leaving Copilot CLI underutilized. This internal competition between Microsoft’s own AI coding tool and a rival product created friction. By removing Claude Code, Microsoft is effectively forcing its developers to use and improve Copilot CLI.
According to an internal note seen by reporters, Microsoft is “encouraging developers to file bug reports and feedback on Copilot CLI ahead of Claude Code being removed.” The hope is that this concentrated feedback will help close the feature gap. However, there are still significant differences between the two tools. Claude Code offers more advanced multi-step reasoning and can handle complex orchestration tasks. Copilot CLI, while improving, still lags in areas like context handling and code generation across large repos.
Financial rationale and the struggle for control
Microsoft’s relationship with Anthropic is multifaceted. In November 2025, Microsoft signed a deal to offer Anthropic’s latest models — Claude Sonnet 4.5, Claude Opus 4.1, and Claude Haiku 4.5 — through its Azure Foundry platform. That commercial agreement remains unaffected by the cancellation of Claude Code licenses. Microsoft continues to use Claude models in its Microsoft 365 Copilot and other products, where they outperform OpenAI’s models for certain tasks. The Claude Code cancellation is therefore a narrow decision, targeting only the standalone developer tool.
However, the financial impact is notable. Microsoft had become one of Anthropic’s largest customers, and the Claude Code licenses represented a recurring expense. Canceling them before the new fiscal year improves Microsoft’s balance sheet without harming its broader AI partnerships. The company is also reported to have considered acquiring Cursor — an AI coding assistant startup — to bolster Copilot, but decided against it due to regulatory scrutiny. Instead, Microsoft is investing internally and working with GitHub to rapidly improve Copilot CLI.
Broader AI strategy at Microsoft
This episode underscores Microsoft’s delicate balancing act between leveraging external AI innovations and controlling its own destiny. The company has heavily invested in OpenAI, but also nurtures relationships with Anthropic and other AI labs. Internally, the use of third-party tools by its own developers risks signaling that Microsoft’s own products are inferior. By forcing engineers onto Copilot CLI, Microsoft aims to demonstrate confidence in its AI stack and collect the data needed to make it world-class.
The move also comes amid reports that Microsoft is revamping its internal AI tooling across the board. The company recently used its own MDASH model to discover security vulnerabilities in its products. It also introduced new features in Windows Update, Edge Copilot, and Xbox applications. The decision to standardize on Copilot CLI is part of a larger effort to create a unified AI ecosystem within Microsoft.
Impact on developer workflows and morale
For many developers inside Microsoft, the loss of Claude Code is a significant change. Some had built workflows around Claude’s code analysis, test generation, and natural language commands. The shift means retraining, adjusting scripts, and potentially losing productivity during the transition. To ease the pain, Microsoft is allowing developers to use Claude models through Copilot CLI — essentially running the same underlying AI but through Microsoft’s wrapper. However, the agentic features differ, and not all Claude capabilities will be available.
There is also a cultural dimension. By cutting off a highly popular tool, Microsoft risks alienating some of its most innovative engineers. The company has long prided itself on being a “developers-first” organization, and dictating which tools must be used contradicts that ethos. On the other hand, Microsoft argues that having a single tool simplifies training, security audits, and compliance. The internal memo from Jha emphasized “shared accountability across GitHub and E+D leadership: to make Copilot CLI the best agentic coding experience for Microsoft engineers.”
Lessons from the Claude Code experiment
The six-month experiment with Claude Code yielded valuable insights. Non-technical employees built prototypes quickly, demonstrating that low-code and AI-assisted development can democratize programming. Managers appreciated the ability to explore ideas without waiting for engineering resources. But the experiment also revealed gaps in Copilot CLI that Microsoft is now racing to fill. The company had considered acquiring Cursor to accelerate its efforts, but instead opted for internal development and closer collaboration with GitHub.
GitHub itself has already shipped significant improvements based on Microsoft’s feedback. New features include better multi-file editing, improved context awareness, and deeper integration with Azure DevOps. Yet Claude Code remains ahead in areas like autonomous code generation and step-by-step reasoning. Microsoft’s developers will now become the primary testers for Copilot CLI, which could accelerate its maturity — or slow down engineering work if the tool is not yet ready.
Wider implications for the AI coding market
Microsoft’s decision signals that even large tech companies struggle to manage internal tool diversity. By consolidating on Copilot CLI, Microsoft is betting that a proprietary, deeply integrated tool will outperform third-party alternatives in the long run. This approach contrasts with companies like Google or Amazon, which allow multiple AI coding assistants to coexist. For Anthropic, losing Microsoft as a customer for Claude Code is a blow, but the company’s overall business with Microsoft remains strong through Azure Foundry and other model licensing deals.
The incident also highlights the delicate nature of partnerships between AI startups and tech giants. Microsoft is both Anthropic’s customer and competitor in the development tools space. The cancellation might strain their relationship, but both parties need each other: Microsoft relies on Anthropic’s models for productivity apps, while Anthropic benefits from Microsoft’s cloud distribution. For now, the Claude Code licenses are gone, but the collaboration continues in other forms.
Source: The Verge News