By ByteBot
GitHub shut down all GitHub App-based Copilot Extensions on November 10, 2025, forcing developers to migrate to the Model Context Protocol (MCP)—an open standard created by competitor Anthropic. The deprecation, announced just six weeks earlier on September 24, gave extension developers a tight timeline to rebuild their integrations from scratch. This wasn’t a migration; GitHub’s documentation explicitly states MCP servers are “architecturally different,” making this a complete replacement.
This move signals a major strategic shift for Microsoft and GitHub. They abandoned their proprietary extension system for an open standard created by a competitor, validating that open protocols beat vendor lock-in in developer tooling.
Why Microsoft Chose a Competitor’s Standard
Microsoft and GitHub joined Anthropic’s MCP steering committee and deprecated their own proprietary Copilot Extensions in favor of MCP. This is rare: a tech giant choosing an open standard created by a competitor over building a closed ecosystem.
The landscape forced their hand. Within months of Anthropic’s MCP launch, OpenAI, Google DeepMind, and AWS adopted it. Developer tools like Cursor, Windsurf, and VS Code shipped MCP support. Fighting this would have fragmented the ecosystem and pushed developers toward competitors.
Moreover, Microsoft learned from past mistakes. Internet Explorer’s proprietary APIs lost to web standards. ActiveX lost to open web technologies. Developers hate vendor lock-in, and MCP’s cross-platform compatibility became table stakes. If GitHub Copilot required proprietary extensions while Cursor and Claude Code worked with open MCP servers, developers would choose the latter.
Furthermore, Microsoft hedged strategically. They back both GitHub Copilot and Anthropic Claude through their Azure partnership. If Copilot loses market share to competitors, Microsoft still benefits from MCP ecosystem growth. As GitHub stated: “An open standard like MCP unlocks more opportunity for the entire ecosystem.”
The Six-Week Migration Scramble
GitHub gave developers a brutal timeline: September 24 announcement, November 3-7 brownout testing, November 10 full sunset. Six weeks to completely rebuild extensions as MCP servers.
The pain came from architectural incompatibility. GitHub’s documentation made it clear: “This is a replacement rather than a migration.” Developers couldn’t port code—they had to learn MCP’s JSON-RPC 2.0 architecture, rebuild integrations from scratch, and deploy within weeks. One developer wrote: “I created a custom Copilot Extension that interacts with my internal document (using RAG). Recently, I read that GitHub is removing Extensions and suggests using MCP instead.” The confusion was real.
Brownout testing in early November exposed gaps. Some developers reported “9+ tools are automatically disabled, breaking automation workflows” with no UI controls to re-enable them. Additionally, others found that GitHub Copilot “supports MCP tools but does not yet support MCP resources,” unlike Claude Desktop’s fuller implementation. Feature parity gaps added friction to an already rushed migration.
What is MCP and Why It Won
The Model Context Protocol solves the “N×M integration problem.” Before MCP, developers built custom connectors for each combination of AI tool and data source. Building a Slack integration for Copilot, Claude, and Cursor meant three separate implementations. MCP provides a universal protocol: build once, works everywhere.
Anthropic describes MCP as “a USB-C port for AI applications—just as USB-C provides a standardized way to connect electronic devices, MCP provides a standardized way to connect AI applications to external systems.” The analogy holds. MCP uses JSON-RPC 2.0 over stdio or HTTP, with three core components: clients (AI tools), servers (integration implementations), and primitives (tools, resources, prompts).
Adoption validates the approach. Anthropic Claude, GitHub Copilot, OpenAI, Google DeepMind, and AWS support MCP. Developer tools—Cursor, Windsurf, Zed, Replit, Codeium, Sourcegraph—ship with MCP integration. Enterprises like Block, Apollo, Klaviyo, and HashiCorp are building MCP servers for their systems. The ecosystem is consolidating fast.
The Death of AI Vendor Lock-In
GitHub’s move to MCP signals a broader industry trend: AI platforms are abandoning vendor lock-in strategies. The competitive landscape made this inevitable. Copilot faces direct competition from Cursor, Windsurf, and Claude Code—all supporting MCP. If Copilot-only extensions existed, developers would switch to competitors offering cross-platform integrations.
This mirrors past platform battles. Java’s “write once, run anywhere” beat Microsoft’s Windows-only .NET for years. Open web standards crushed proprietary ActiveX controls. Consequently, developers prefer tools that don’t lock them into a single vendor, and the market rewards companies that embrace openness over those that don’t.
Microsoft’s strategy here is smart, even if painful for extension developers. By backing MCP, they position Azure as the cloud platform for MCP-integrated AI agents, regardless of which AI tool wins. Windows 11 will expose app functionalities as MCP servers. Microsoft even developed an official C# SDK for MCP. They’re betting on the ecosystem, not just Copilot.
The short-term pain for developers—rushed timelines, broken workflows, emergency rebuilds—serves a long-term benefit. MCP eliminates integration lock-in. Developers can choose the best AI tool without losing access to their data sources. Competition shifts from “who has the most integrations” to “who has the best AI capabilities.” That’s a healthier market.
Key Takeaways
- MCP is rapidly becoming the universal standard for AI agent integration across Copilot, Claude, Cursor, Windsurf, and more
- Open standards beat proprietary lock-in in developer tools—Microsoft learned this lesson from Internet Explorer and ActiveX
- GitHub’s six-week migration timeline was painful, but architectural incompatibility made clean breaks necessary
- Microsoft hedged strategically by backing both Copilot and Anthropic’s MCP through Azure partnerships
- Expect accelerated MCP ecosystem growth in 2026 as more platforms and tools standardize on the protocol
