Filesystem MCP Server
Node.js server implementing Model Context Protocol (MCP) for filesystem operations with comprehensive permission controls, allowing secure file and directory manipulation with granular access restrictions.
README
Filesystem MCP Server
Node.js server implementing Model Context Protocol (MCP) for filesystem operations with comprehensive permission controls and enhanced functionality.
Features
- Granular permission controls (read-only, full access, or specific operation permissions)
- Secure file operations within allowed directories
- File operations:
- Read/write/modify files
- Create/list/delete directories
- Move files/directories
- Search files by name or extension
- Get file metadata
- Directory operations:
- Tree view of directory structures
- Recursive operations with exclusion patterns
- Utility functions:
- XML to JSON conversion
- Multiple file operations in one call
- Advanced file editing with pattern matching
- Security features:
- Symlink control
- Path validation
- Sandboxed operations
Note: The server will only allow operations within directories specified via args and according to the configured permissions.
API
Resources
file://system: File system operations interface
Tools
-
read_file
- Read complete contents of a file
- Input:
path(string) - Reads complete file contents with UTF-8 encoding
-
read_multiple_files
- Read multiple files simultaneously
- Input:
paths(string[]) - Failed reads won't stop the entire operation
-
create_file
- Create a new file with content
- Inputs:
path(string): File locationcontent(string): File content
- Fails if file already exists
- Requires
createpermission
-
modify_file
- Modify an existing file with new content
- Inputs:
path(string): File locationcontent(string): New file content
- Fails if file doesn't exist
- Requires
editpermission
-
edit_file
- Make selective edits using pattern matching and formatting
- Features:
- Line-based and multi-line content matching
- Whitespace normalization with indentation preservation
- Multiple simultaneous edits with correct positioning
- Indentation style detection and preservation
- Git-style diff output with context
- Preview changes with dry run mode
- Inputs:
path(string): File to editedits(array): List of edit operationsoldText(string): Text to search for (exact match)newText(string): Text to replace with
dryRun(boolean): Preview changes without applying (default: false)
- Returns detailed diff for dry runs, otherwise applies changes
- Requires
editpermission - Best Practice: Always use dryRun first to preview changes
-
create_directory
- Create new directory or ensure it exists
- Input:
path(string) - Creates parent directories if needed
- Succeeds silently if directory exists
- Requires
createpermission
-
list_directory
- List directory contents with [FILE] or [DIR] prefixes
- Input:
path(string) - Returns detailed listing of files and directories
-
directory_tree
- Get recursive tree view of directory structure
- Input:
path(string) - Returns JSON structure with files and directories
- Each entry includes name, type, and children (for directories)
-
move_file
- Move or rename files and directories
- Inputs:
source(string): Source pathdestination(string): Destination path
- Fails if destination exists
- Works for both files and directories
- Requires
movepermission
-
delete_file
- Delete a file
- Input:
path(string) - Fails if file doesn't exist
- Requires
deletepermission
-
delete_directory
- Delete a directory
- Inputs:
path(string): Directory to deleterecursive(boolean): Whether to delete contents (default: false)
- Fails if directory is not empty and recursive is false
- Requires
deletepermission
-
search_files
- Recursively search for files/directories
- Inputs:
path(string): Starting directorypattern(string): Search patternexcludePatterns(string[]): Exclude patterns (glob format supported)
- Case-insensitive matching
- Returns full paths to matches
-
find_files_by_extension
- Find all files with specific extension
- Inputs:
path(string): Starting directoryextension(string): File extension to findexcludePatterns(string[]): Optional exclude patterns
- Case-insensitive extension matching
- Returns full paths to matching files
-
get_file_info
- Get detailed file/directory metadata
- Input:
path(string) - Returns:
- Size
- Creation time
- Modified time
- Access time
- Type (file/directory)
- Permissions
-
get_permissions
- Get current server permissions
- No input required
- Returns:
- Permission flags (readonly, fullAccess, create, edit, move, delete)
- Symlink following status
- Number of allowed directories
-
list_allowed_directories
- List all directories the server is allowed to access
- No input required
- Returns array of allowed directory paths
-
xml_to_json
- Convert XML file to JSON format
- Inputs:
xmlPath(string): Source XML filejsonPath(string): Destination JSON fileoptions(object): Optional settingsignoreAttributes(boolean): Skip XML attributes (default: false)preserveOrder(boolean): Maintain property order (default: true)format(boolean): Pretty print JSON (default: true)indentSize(number): JSON indentation (default: 2)
- Requires
readpermission for XML file - Requires
createoreditpermission for JSON file
-
xml_to_json_string
- Convert XML file to JSON string
- Inputs:
xmlPath(string): Source XML fileoptions(object): Optional settingsignoreAttributes(boolean): Skip XML attributes (default: false)preserveOrder(boolean): Maintain property order (default: true)
- Requires
readpermission for XML file - Returns JSON string representation
-
xml_query
- Query XML file using XPath expressions
- Inputs:
path(string): Path to the XML filequery(string, optional): XPath query to executestructureOnly(boolean, optional): Return only tag structuremaxBytes(number, optional): Maximum bytes to read (default: 1MB)includeAttributes(boolean, optional): Include attribute info (default: true)
- XPath examples:
- Get all elements:
//tagname - Get elements with specific attribute:
//tagname[@attr="value"] - Get text content:
//tagname/text()
- Get all elements:
- Memory efficient for large XML files
- Returns JSON representation of query results or structure
-
xml_structure
- Analyze XML structure without reading entire file
- Inputs:
path(string): Path to the XML filedepth(number, optional): How deep to analyze (default: 2)includeAttributes(boolean, optional): Include attribute analysismaxBytes(number, optional): Maximum bytes to read (default: 1MB)
- Returns statistical information about elements, attributes, and structure
- Useful for understanding large XML files before detailed analysis
Permissions & Security
The server implements a comprehensive security model with granular permission controls:
Directory Access Control
- Operations are strictly limited to directories specified during startup via
args - All operations (including symlink targets) must remain within allowed directories
- Path validation ensures no directory traversal or access outside allowed paths
Permission Flags
- --readonly: Enforces read-only mode, overriding all other permission flags
- --full-access: Enables all operations (create, edit, move, delete)
- Individual permission flags (require explicit enabling unless --full-access is set):
- --allow-create: Allow creation of new files and directories
- --allow-edit: Allow modification of existing files
- --allow-move: Allow moving/renaming files and directories
- --allow-delete: Allow deletion of files and directories
Default Behavior: If no permission flags are specified, the server runs in read-only mode. To enable any write operations, you must use either --full-access or specific --allow-* flags.
Symlink Handling
- By default, symlinks are followed (both link and target must be in allowed directories)
- --no-follow-symlinks: Disable symlink following (operations act on the link itself)
Usage with Claude Desktop and Cursor
Add appropriate configuration to either claude_desktop_config.json (for Claude Desktop) or .cursor/mcp.json (for Cursor):
Cursor Configuration
In .cursor/mcp.json:
{
"mcpServers": {
"my-filesystem": {
"command": "node",
"args": [
"/path/to/mcp-filesystem/dist/index.js",
"~/path/to/allowed/directory",
"--full-access"
]
}
}
}
Docker Configuration
For Claude Desktop with Docker:
{
"mcpServers": {
"filesystem": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"--mount", "type=bind,src=/Users/username/Desktop,dst=/projects/Desktop",
"--mount", "type=bind,src=/path/to/other/allowed/dir,dst=/projects/other/allowed/dir,ro",
"--mount", "type=bind,src=/path/to/file.txt,dst=/projects/path/to/file.txt",
"mcp/filesystem",
"--readonly", // For read-only access
"--no-follow-symlinks", // Optional: prevent symlink following
"/projects"
]
}
}
}
NPX Configuration
For either Claude Desktop or Cursor with NPX:
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-filesystem",
"--full-access", // For full read/write access
"/Users/username/Desktop",
"/path/to/other/allowed/dir"
]
}
}
}
Permission Flag Examples
You can configure the server with various permission combinations:
"args": [
"/path/to/mcp-filesystem/dist/index.js",
"~/path/to/allowed/directory",
"--readonly" // Read-only mode
]
"args": [
"/path/to/mcp-filesystem/dist/index.js",
"~/path/to/allowed/directory",
"--full-access", // Full read/write access
"--no-follow-symlinks" // Don't follow symlinks
]
"args": [
"/path/to/mcp-filesystem/dist/index.js",
"~/path/to/allowed/directory",
"--allow-create", // Selective permissions
"--allow-edit" // Only allow creation and editing
]
Note: --readonly takes precedence over all other permission flags, and --full-access enables all operations unless --readonly is specified.
Multiple Directories and Permissions
When specifying multiple directories, permission flags apply globally to all directories:
"args": [
"/path/to/mcp-filesystem/dist/index.js",
"~/first/directory", // Both directories have the same
"~/second/directory", // permission settings (read-only)
"--readonly"
]
If you need different permission levels for different directories, create multiple server configurations:
{
"mcpServers": {
"readonly-filesystem": {
"command": "node",
"args": [
"/path/to/mcp-filesystem/dist/index.js",
"~/sensitive/directory",
"--readonly"
]
},
"writeable-filesystem": {
"command": "node",
"args": [
"/path/to/mcp-filesystem/dist/index.js",
"~/sandbox/directory",
"--full-access"
]
}
}
}
Command Line Examples
- Read-only access:
npx -y @modelcontextprotocol/server-filesystem --readonly /path/to/dir
- Full access:
npx -y @modelcontextprotocol/server-filesystem --full-access /path/to/dir
- Specific permissions:
npx -y @modelcontextprotocol/server-filesystem --allow-create --allow-edit /path/to/dir
- No symlink following:
npx -y @modelcontextprotocol/server-filesystem --full-access --no-follow-symlinks /path/to/dir
Build
Docker build:
docker build -t mcp/filesystem -f src/filesystem/Dockerfile .
License
This MCP server is licensed under the MIT License. This means you are free to use, modify, and distribute the software, subject to the terms and conditions of the MIT License. For more details, please see the LICENSE file in the project repository.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。