VS Code拡張機能をClaude Codeでゼロから作って、マーケットプレイスに公開したら1時間で10件ダウンロードされた話 VS Code拡張機能をClaude Codeでゼロから作って、マーケットプレイスに公開したら1時間で10件ダウンロードされた話

A nagging but familiar frustration in software development — wanting to peek inside a ZIP file without actually extracting it — became the seed of a small but telling experiment in AI-assisted tooling. A developer used Claude Code, Anthropic's agentic coding assistant, to build a VS Code extension from scratch that renders ZIP archive contents as a navigable tree view directly inside the editor. The result was published to the Visual Studio Marketplace, and within an hour it had collected ten downloads.

VS Code's default behavior when encountering archive files is to either render raw binary output or hand off to an external application. Neither option is frictionless. The conventional workaround — extract, inspect, delete — is a minor but repeated tax on developer attention. The author identified this gap and decided to close it with a custom extension rather than waiting for someone else to do it.

Claude Code served as the primary development partner throughout the project. The tool supports an iterative, conversational workflow where developers describe intent and Claude Code generates or refines code in response. Building a VS Code extension normally requires familiarity with TypeScript and the VS Code Extension API — a non-trivial learning curve for developers venturing outside their usual stack. By delegating much of the scaffolding and implementation detail to Claude Code while retaining review and directional control, the author was able to reach a working product in a compressed timeframe. This human-in-the-loop pattern, where AI handles boilerplate and first drafts while the developer steers, is increasingly central to how small utilities get built quickly.

Publishing to the Visual Studio Marketplace involves registering a Publisher account through Azure DevOps, packaging the extension with the vsce CLI tool, and navigating Microsoft's review pipeline. The infrastructure is well-documented but still represents a meaningful step for first-time publishers. The fact that the author completed both development and publication in a single session underscores how much friction AI coding tools can absorb.

The ten downloads within the first hour are a small number in absolute terms, but they carry signal. The Visual Studio Marketplace hosts well over 40,000 extensions, making discoverability a genuine challenge. Early organic traction on a niche utility suggests real latent demand — developers who have quietly endured the same workflow friction and are quick to adopt a targeted fix when it surfaces. Niche-but-precise tools have historically punched above their weight in extension ecosystems precisely because they solve specific pain with minimal configuration overhead.

This story fits into a broader trend worth watching. Tools like Claude Code, GitHub Copilot Workspace, and Cursor are lowering the barrier for developers to ship polished, distributable software from an idea in an afternoon. The practical consequence is a coming wave of small, highly targeted developer tools — extensions, CLI utilities, integrations — authored with AI assistance and released to public registries. Whether that volume ultimately improves signal-to-noise in ecosystems like the VS Code Marketplace is an open question, but cases like this one suggest the tooling to support rapid idea-to-publish cycles is already mature enough to use in production.