Filesystem MCP Server (@shtse8/filesystem-mcp)

@shtse8

リポジトリホームページ

基本情報

カテゴリ

ファイルとストレージ

ライセンス

MIT

トランスポート

stdio

作者

shtse8

Filesystem MCP Server について

📁 Secure, efficient MCP filesystem server - token-saving batch operations with project root confinement

#nodejs #npm #typescript #filesystem #mcp #cline #ai-agent #claude-ai #model-context-protocol #roo-code

Filesystem MCP Server の設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "filesystem-mcp": {
      "command": "npx",
      "args": [
        "@sylphx/filesystem-mcp"
      ],
      "name": "Filesystem (npx)"
    }
  }
}

Filesystem MCP Server のツール

ツール一覧はまだありません

この掲載には表示できるツール一覧がまだありません。サーバーにツールがないという意味ではありません。対応機能はプロジェクトのドキュメントをご確認ください。

上のボタンで一時的なサンドボックスでこのサーバーを実行し、ツール一覧を取得できます。

概要

What is @shtse8/filesystem-mcp?

@Shtse8/filesystem-mcp is a Node.js server implementing the Model Context Protocol (MCP) that provides AI agents with secure, token‑optimized filesystem operations. It confines all file access to the project root, supports batch processing of multiple files per request, and uses Zod validation. It is designed for developers using AI coding assistants like Claude or Cline.

How to use @shtse8/filesystem-mcp?

Configure the server in your MCP host (e.g., Cline, VSCode) using npx or bunx with the command @shtse8/filesystem-mcp. Alternatively, run the official Docker image, mounting your project directory to /app. The server uses its current working directory as the project root; ensure your host sets cwd appropriately.

Key features of @shtse8/filesystem-mcp

Batch operations – read, write, edit multiple files in one request
Token‑optimized – reduces round trips and overhead (50‑70% less token usage)
Project root confinement – prevents path traversal outside the working directory
Zod validation – all tool arguments validated for safety
Detailed per‑item error reporting for batch operations
11+ tools including read, write, edit, search, move, copy, delete, chmod, chown

Use cases of @shtse8/filesystem-mcp

AI agents reading, editing, and searching project files during development
Batch refactoring across multiple files in a codebase
Safe automated file generation, project setup, and directory management
Integration with code assistants (Claude, Cline) for direct filesystem access

よくある質問

How does it differ from using shell commands?

It replaces individual shell commands (one file per operation, high token cost and latency) with direct API calls that batch 10+ files per request. This reduces token usage by 50–70% and latency 5–10×, with per‑item success/failure reporting instead of stderr parsing.

What security measures are in place?

All operations are confined to the server’s current working directory (project root). Path traversal is prevented. Built‑in `chmod` and `chown` tools allow permission control. Arguments are validated with Zod schemas.

Which tools support batch operations?

Most tools support batch processing: `read_content`, `write_content`, `edit_file`, `search_files`, `replace_content`, `stat_items`, `create_directories`, `delete_items`, `move_items`, `copy_items`, `chmod_items`, `chown_items`. Only `list_files` operates on a single path.

What runtime is required?

Node.js or Bun. The server is distributed as an npm package (`@sylphx/filesystem-mcp`) and a Docker image (`sylphx/filesystem-mcp`). It is built with TypeScript and uses pnpm for development.

How do I install and configure it?

Use `npx @shtse8/filesystem-mcp` or `bunx @shtse8/filesystem-mcp` in your MCP host’s `mcpServers` config. Alternatively, run `docker run -i --rm -v /path/to/project:/app sylphx/filesystem-mcp:latest`. Ensure the server’s `cwd` points to your project root.

「ファイルとストレージ」の他のコンテンツ

HTMLvault

AI tools (Claude, ChatGPT, Gemini, Grok, etc.) generate HTML constantly, from reports to landing pages to prototypes, but there's no easy way to share it or know if anyone looked. HTMLvault turns any HTML into a clean, t

S3 MCP Server

samuraikun

MCP server to integrate AWS S3 and LLM

PDF Reader MCP Server (@shtse8/pdf-reader-mcp)

shtse8

📄 The PDF intelligence layer for AI agents — Agent Document Twin, evidence-first extraction, visual crops, OCR provenance, trust reports, and benchmark-gated releases. MCP server for Claude, Cursor, VS Code, and any MCP client.

810