GitHub Copilot の利用量を可視化し、残り続ける履歴を安全に削除する VS Code 拡張を作ったA developer built a VS Code extension that visualizes GitHub Copilot usage…

GitHub Copilot has become a fixture in many developers' daily workflows, yet two mundane problems tend to go unaddressed: knowing how much you actually use it, and managing the chat history that quietly piles up over time. A developer has built a Visual Studio Code extension that tackles both, pairing a usage dashboard with a tool for safely clearing accumulated Copilot chat records, according to a post published on the Japanese developer platform Zenn. The appeal is practical rather than flashy: it folds usage visibility and storage maintenance into a single interface.

The extension appears aimed at individual users rather than team administrators. GitHub already offers organization- and enterprise-level reporting through its own dashboards and a Copilot metrics API, but individuals on personal or Business plans often have limited insight into their own patterns of use. By surfacing usage statistics directly inside the editor, the extension lets users gauge their activity without leaving VS Code or signing into a separate web console. That kind of at-a-glance awareness has grown more relevant as GitHub has moved parts of Copilot toward usage-based billing, where certain "premium" requests against more capable models are metered, making it useful to understand consumption before it translates into cost.

The second half of the tool addresses storage. Copilot Chat retains conversation history so that context can persist across sessions, but that data keeps growing, and over months of heavy use it can occupy noticeable disk space and clutter the interface. The extension provides a way to review and delete this history in a controlled manner, which the developer frames as a "safe" cleanup. That distinction matters, because manually editing the files where VS Code and its extensions keep state can risk corrupting a workspace or removing data unintentionally. A purpose-built routine that targets only the relevant records is less error-prone than digging through storage directories by hand.

Technically, VS Code extensions are typically written in TypeScript or JavaScript and distributed through the Visual Studio Marketplace or the Open VSX registry. They run within the editor's extension host and can access defined storage locations, invoke the editor's APIs, and render custom views such as webview-based panels for dashboards. Chat and workspace state in VS Code is commonly held in local databases and JSON files under the user's profile directory. It is likely that the extension reads from and manages these locations to compile its statistics and perform its cleanup, though the precise data sources and the exact metrics it exposes are best confirmed against the extension's own documentation, since the internal formats used by Copilot are not part of a stable public contract and can change between releases.

The project also fits a broader industry context. AI coding assistants have proliferated, with alternatives such as Cursor, Sourcegraph's Cody, Amazon Q Developer, and Windsurf competing alongside Copilot, and each accumulates its own logs, caches, and conversation data as it becomes embedded in everyday coding. As these tools mature, a secondary ecosystem of companion utilities has emerged to help users observe, tune, and maintain them. Extensions that track token or request usage, manage prompt libraries, or clean up generated artifacts reflect the same instinct on display here: the assistant is useful, but it needs housekeeping.

For readers considering the tool, a few caveats are worth keeping in mind. Individually built extensions can vary in maintenance and support, and because this one interacts with Copilot-related data, it is prudent to review its permissions and understand what it deletes before running a cleanup, ideally after confirming that any important chat context has been preserved elsewhere. Users on managed or enterprise devices may also be subject to organizational policies governing which extensions can be installed.

Even so, the underlying idea is a sensible one. Rather than treating usage analytics and storage hygiene as separate chores handled by separate tools, the extension consolidates them where the work already happens, inside the editor. As AI assistance becomes a standard layer of the development environment, small utilities that make its footprint visible and manageable are likely to remain valuable, and this project is a concrete example of a developer scratching a common itch and sharing the result with the wider community.