File Server MCP
# File Server MCP
A Model Context Protocol (MCP) server that exposes documents from specific folders through configurable access.
## Features
- **Configurable Folders**: Expose specific folders like Documents, Reports, Desktop, etc.
- **File Type Filtering**: Only allow specific file extensions for security
- **Size Limits**: Configurable maximum file size limits
- **Tools Available**:
- `list_files`: List files in a configured folder with optional pattern matching
- `read_file`: Read the content of a specific file
- `search_files`: Search for files containing specific text
- **Resources**: Access files through MCP resource URIs like `file://documents/filename.txt`
## Configuration
The server can be configured through environment variables or the `config.json` file:
### Default Folders
- `documents`: C:\Users\bill\Documents
- `reports`: C:\Support
- `desktop`: C:\Users\bill\Desktop
- `backup_scripts`: C:\support
### Environment Variables
- `FILE_SERVER_CONFIG`: JSON string with custom configuration
- `DOCUMENTS_FOLDER`: Override documents folder path
- `REPORTS_FOLDER`: Override reports folder path
- `DESKTOP_FOLDER`: Override desktop folder path
### Configuration File
Edit `config.json` to customize:
```json
{
"folders": {
"documents": "C:\\Users\\bill\\Documents",
"reports": "C:\\Support",
"desktop": "C:\\Users\\bill\\Desktop",
"backup_scripts": "C:\\support"
},
"allowedExtensions": [".txt", ".md", ".pdf", ".docx", ".xlsx", ".csv", ".json", ".xml", ".log", ".ps1"],
"maxFileSize": 10485760
}
```
## Installation
### Prerequisites
1. Install Node.js (https://nodejs.org/)
2. Install npm dependencies:
```bash
npm install
```
### Build
```bash
npm run build
```
### MCP Configuration
Add to your MCP settings file (`C:\Users\bill\AppData\Roaming\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json`):
```json
{
"mcpServers": {
"file-server": {
"command": "node",
"args": ["C:/Users/bill/Desktop/file-server/build/index.js"],
"env": {
"FILE_SERVER_CONFIG": "{\"folders\":{\"documents\":\"C:\\\\Users\\\\bill\\\\Documents\",\"reports\":\"C:\\\\Support\",\"desktop\":\"C:\\\\Users\\\\bill\\\\Desktop\",\"backup_scripts\":\"C:\\\\support\"}}"
}
}
}
}
```
## Usage Examples
Once installed, you can use the file server through MCP tools:
### List Files
```
Use the list_files tool with folder="documents" to see all files in the Documents folder
```
### Read a File
```
Use the read_file tool with folder="reports" and filename="backup-report.csv" to read a specific file
```
### Search Files
```
Use the search_files tool with folder="backup_scripts" and searchText="SQLite" to find scripts containing "SQLite"
```
### Access Resources
```
Access file://documents/ to get a list of files in Documents
Access file://reports/backup-report.csv to read a specific file
```
## Security Features
- Only configured folders are accessible
- File type restrictions through allowedExtensions
- File size limits to prevent large file access
- Path traversal protection
- Read-only access (no file modification)
## Folder Structure
```
file-server/
├── package.json # Node.js package configuration
├── tsconfig.json # TypeScript configuration
├── config.json # Server configuration
├── README.md # This file
├── src/
│ └── index.ts # Main server implementation
└── build/ # Compiled JavaScript (after npm run build)
└── index.js # Executable MCP server
```
## Development
```bash
# Watch mode for development
npm run dev
# Build for production
npm run build
# Test the server
npm start
TDQS
Scored across 3 tools
Each tool targets a distinct operation: listing directory entries, reading one file's content, and searching for files by text. There is little risk of an agent selecting the wrong tool.
All three tools follow a consistent verb_noun snake_case pattern (list_files, read_file, search_files). The singular/plural noun forms are conventional and do not create ambiguity.
Three tools is minimal but reasonable for a narrowly scoped read-only file access server. It is slightly lean, but every tool covers a distinct and useful operation.
Despite the 'File Server' name, there are no create, write, update, move, or delete operations, leaving file modification workflows impossible. Agents can only read and search, so the surface is notably incomplete for general file management.