GitHub Copilot in Visual Studio Code — 2026年5月リリースまとめ (v1.120〜v1.123) GitHub Copilot in Visual Studio Code, May releases

GitHub Copilot for Visual Studio Code maintained its weekly release cadence throughout May 2026, shipping four stable updates — v1.120 through v1.123 — that collectively push the tool further along its trajectory from code autocomplete toward a fully autonomous development assistant.

The headline item across this release window is the general availability of Agents mode. After an extended preview period during which GitHub gathered feedback and hardened stability, Agents mode is now production-ready for all Copilot subscribers. Unlike traditional inline suggestions, Agents mode enables Copilot to take multi-step actions: editing files across a project, running terminal commands, querying external APIs, and iterating on its own output — all within a single coherent workflow. The shift to GA signals that GitHub considers the feature stable enough for everyday professional use, not just experimentation.

The broader context here matters. GitHub has spent the past year distributing Copilot across multiple surfaces — CLI, mobile, GitHub.com, and the IDE — but VS Code remains the flagship integration. Microsoft's tight co-ownership of both VS Code and GitHub allows the two teams to wire Copilot deeply into the editor's extension APIs, language server protocol infrastructure, and debug adapter layer in ways that standalone AI editors like Cursor or Windsurf must approximate from the outside. That architectural advantage becomes increasingly meaningful as agentic features require richer hooks into the development environment.

The competitive landscape has grown crowded. JetBrains now bundles AI Assistant across its full IDE lineup, and AI-native editors continue to attract developers who want AI woven into every interaction from day one. Yet VS Code's installed base — conservatively estimated in the tens of millions — gives Copilot a distribution moat that newcomers will struggle to overcome regardless of feature parity.

Weekly stable releases are themselves a strategic signal. They indicate a mature CI/CD pipeline and enough test coverage to ship frequently without destabilizing existing users. For developers, it means new capabilities arrive on a predictable schedule, lowering the cost of staying current.

Looking ahead, the GA of Agents mode is likely a stepping stone rather than an endpoint. Integration with the Model Context Protocol (MCP) — which allows language models to consume structured context from external tools and services — is expected to deepen, potentially letting Copilot pull in project documentation, issue trackers, or CI results as first-class inputs. Whether that vision materializes on the timescale GitHub implies remains to be seen, but the May releases suggest the team is executing with consistency.