MCP File System Server

@MarcusJellinghaus

代码仓库

基本信息

分类

文件与存储

许可证

MIT

传输方式

stdio

作者

MarcusJellinghaus

关于 MCP File System Server

MCP Workspace Server: A secure Model Context Protocol server providing file, git, and GitHub tools for AI assistants within a sandboxed project directory.

#open-source #filesystem #mcp #file-operations #ai-assistant #claude-ai #claude-desktop #model-context-protocol #mcp-server

MCP File System Server 配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "mcp_server_filesystem": {
      "command": "python",
      "args": [
        "-m",
        "venv",
        "venv"
      ]
    }
  }
}

MCP File System Server 工具

Lists files and directories in the project directory

Reads the contents of a file

Creates or overwrites files atomically

Adds content to existing files

Removes files from the filesystem

Makes selective edits using exact string matching

Moves or renames files and directories

Lists available reference projects

Lists files in a reference project

Reads files from reference projects

概览

What is MCP File System Server?

A Model Context Protocol (MCP) server that enables AI assistants (e.g., Claude via Claude Desktop) to read, write, edit, delete, and manage files within a specified project directory. All file operations are securely contained to that directory, with optional read‑only reference projects for additional context.

How to use MCP File System Server?

Install by cloning the repository and running pip install -e . in the project directory. Start the server with mcp-workspace --project-dir /path/to/project and optional arguments for reference projects, logging level, log file path, and file size limit. For Claude Desktop integration, add the server configuration to claude_desktop_config.json with the mcp-workspace command and required --project-dir argument.

Key features of MCP File System Server

Directory listing with gitignore filtering and exclusion of .git folders
Reading files with optional line ranges and line numbers
Atomic file creation/overwrite with save_file
Append content to existing files with append_file
Precise text editing using exact string matching (edit_file)
Move or rename files/directories within the project
Read‑only access to multiple external reference projects
Structured logging in both human‑readable and JSON formats

Use cases of MCP File System Server

Read existing project files for analysis or summarization
Generate and write new files based on natural language descriptions
Make targeted edits to fix bugs or update code without rewriting entire files
Delete temporary or unwanted files from the project directory
Reference documentation, examples, or shared libraries from other directories

常见问题

What runtime dependencies does the MCP File System Server require?

Python 3 and pip are required. The server can be installed from source by cloning the repository and running `pip install -e .`.

How does the MCP File System Server ensure security?

All file operations are restricted to the main project directory specified with `--project-dir`. Reference projects are strictly read‑only, and all paths are validated to prevent directory traversal attacks. Gitignore filtering is applied automatically.

How can I integrate the MCP File System Server with Claude Desktop?

Add an entry to `claude_desktop_config.json` under `mcpServers` with the command `mcp-workspace` and appropriate arguments, including the required `--project-dir` path. Restart Claude Desktop to apply the change.

Can the MCP File System Server access additional read‑only directories?

Yes. Use the `--reference-project` argument (repeatable) in the format `name=/path/to/dir` to provide read‑only access to extra codebases, documentation, or configuration files.

Is there a file size limit when reading or editing files?

The server provides a `check_file_size` tool with a default line limit of 600 lines. This limit can be customized with the `--file-size-limit` argument.

文件与存储分类下的更多 MCP 服务器

drop.media

Upload once. Find, use and share across your AI workflows. drop.media is one media library for you and every AI agent you use: files added from Claude, ChatGPT, Cursor, the terminal or the web can be found, reused and sh