Void MCP Filesystem Server

A secure, sandboxed MCP (Model Context Protocol) server that provides Void Editor with safe filesystem access. Enable your AI to create, read, modify, and delete files within your project—all with comprehensive security controls.

Features

✅ File Operations - Create, read, write, delete, and move files
✅ Directory Management - Create and list directories
✅ Search Capability - Find text across multiple files
✅ Security Sandboxing - Restricted paths, blocked extensions, size limits
✅ Easy Setup - Simple configuration with JSON
✅ Error Handling - Clear feedback on operations

Quick Start

Prerequisites

• Python 3.10 or higher
• Void Editor
• pip (Python package manager)

Installation

1. Clone the repository

git clone https://github.com/yourusername/void-mcp-filesystem-server.git
cd void-mcp-filesystem-server

2. Create a virtual environment

python3 -m venv .venv
source .venv/bin/activate  # On macOS/Linux
# or on Windows:
# .venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Create configuration file (optional)

cp mcp_config.example.json mcp_config.json
# Edit mcp_config.json with your settings

Configure Void Editor

Option 1: Via Void Settings UI

1. Open Void Editor
2. Go to Settings/Preferences
3. Find the MCP Configuration section
4. Add your server:
   • Command: /path/to/.venv/bin/python3
   • Args: ["/full/path/to/mcp_server.py"]
   • Working Directory: Your project root

Option 2: Edit MCP Config File

Create or edit the MCP servers config file:

macOS/Linux: ~/.config/void/mcp_servers.json
Windows: %APPDATA%\void\mcp_servers.json

{
  "mcpServers": {
    "filesystem": {
      "command": "/absolute/path/to/.venv/bin/python3",
      "args": ["/absolute/path/to/mcp_server.py"],
      "cwd": "/your/project/directory"
    }
  }
}

Important: Use absolute paths for both the Python executable and script.

Add System Prompt to Void

In Void's settings, find the prompt/instructions section and add:

You have access to filesystem tools via the MCP (Model Context Protocol) server. Use them to create, read, modify, and delete files as requested by the user.

Available tools:
- read_file(path: str) - Read file contents
- write_file(path: str, content: str) - Write or overwrite a file
- create_file(path: str, content: str = "") - Create a new file (fails if exists)
- edit_file(file_path: str, old_string: str, new_string: str, replace_all: bool = False) - Precisely edit file by replacing exact text
- delete_file(path: str) - Delete a file
- list_directory(path: str = ".") - List directory contents
- create_directory(path: str) - Create a directory
- move_file(source: str, destination: str) - Move or rename a file
- search_in_files(search_term: str, directory: str = ".", file_pattern: str = "*.py") - Search for text in files

When calling these tools, use the exact function name and parameters. Paths can be relative or absolute (if within project). Always use forward slashes (/) in paths.

Examples:
- Create: create_file("src/utils.py", "def helper():\n    pass")
- Read: read_file("src/main.py")
- Edit: edit_file("config.py", "DEBUG = True", "DEBUG = False")
- Modify: For complete rewrites, use read_file() then write_file()
- Search: search_in_files("TODO", "src", "*.py")

Usage

Once configured, ask your AI to perform file operations:

• "Create a new Python file at src/handlers.py with basic structure"
• "Read config.json and explain its structure"
• "In main.py, change the DEBUG variable from True to False"
• "Add a docstring to the calculate function in utils.py"
• "Search for all TODO comments in the codebase"
• "Create a tests/ directory and add a test file"

Configuration

Edit mcp_config.json to customize security settings:

{
  "allowed_root": "/absolute/path/to/your/project",
  "additional_blocked": [
    "secrets",
    "credentials"
  ],
  "additional_extensions": [
    ".vue",
    ".svelte"
  ],
  "max_file_size_mb": 10
}

Configuration Options

• allowed_root - Project directory where operations are restricted (default: current directory)
• additional_blocked - Extra patterns to block (beyond defaults like .git, .env, .ssh)
• additional_extensions - Additional file extensions to allow
• max_file_size_mb - Maximum file size in megabytes (default: 10)

Security

The server implements multiple layers of protection:

Path Restrictions

• All operations confined to allowed_root directory
• No directory traversal attacks (../../../etc/passwd blocked)
• Uses Path.resolve() for absolute path validation

File Blocking

• Blocks sensitive files: .env, .key, .pem, id_rsa, etc.
• Blocks directories: .git, .ssh, node_modules, __pycache__
• Only allows specific text file extensions

Size Limits

• Default 10MB file size limit
• Prevents memory issues from large file operations

No Shell Access

• No subprocess or shell command execution
• Safe even if AI behaves unexpectedly

Testing

Run the included test suite to verify all functionality:

1. Create a test directory: mkdir test_workspace && cd test_workspace
2. Configure MCP to use this directory
3. Follow the test cases in TEST_CASES.md

Each test verifies a specific tool and includes security tests.

Troubleshooting

Server Won't Start

# Check Python version
python3 --version  # Must be 3.10+

# Check mcp is installed
pip list | grep mcp

# Try running manually
python3 mcp_server.py

Void Can't Connect

• Verify paths in MCP config are absolute
• Check Python interpreter path is correct
• Ensure working directory exists
• Check Void's developer console for errors

LLM Won't Use Tools

• Verify system prompt was added to Void settings
• Check that MCP server appears connected (look for tool icon)
• Try restarting Void
• Confirm model supports tool use (Claude, GPT-4, Command-R+ work best)

Access Denied Errors - "Path outside allowed directory"

⚠️ Known Void Issue: Void currently restricts MCP servers to the Void installation directory, ignoring the cwd setting in mcp.json. This is a bug in Void's MCP channel implementation (see Void GitHub Issue).

Workaround: Place your project inside Void's installation directory:

# Move your project to Void's directory
mv /your/project/path /home/edible/Tools/Void/your_project

Then update mcp_config.json:

{
  "allowed_root": "/home/edible/Tools/Void/your_project"
}

Note: This is a temporary workaround. The proper fix requires changes to Void's MCP implementation to respect the configured cwd.

Other Access Denied Errors

• Check file/directory is within allowed_root
• Verify file extension is in allowed list
• Check if path contains blocked patterns (.git, .env, etc.)
• Review mcp_config.json settings

Advanced Usage

Custom Tool Prompts

In mcp_server.py, the system prompt is automatically provided via the @mcp.prompt() decorator. Modify this function to customize tool instructions for your use case.

Extending Tools

Add new tools by creating functions with the @mcp.tool() decorator:

@mcp.tool()
def count_lines(path: str) -> str:
    """Count lines in a file"""
    with open(path) as f:
        return str(len(f.readlines()))

Adding Resources

Expose read-only data via @mcp.resource():

@mcp.resource("custom://data")
def get_custom_data() -> str:
    """Provide custom context data"""
    return "Custom data here"

Limitations

• No symbolic link support
• No binary file operations
• No arbitrary command execution (intentional for security)
• 10MB file size limit (configurable)
• Operations slower than native filesystem access (network overhead)

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Testing & Validation

Before submitting PRs:

• Run all test cases in TEST_CASES.md
• Test security constraints (path traversal, blocked extensions)
• Verify error messages are clear
• Check that operations work with different LLM models

License

This project is licensed under the MIT License - see LICENSE file for details.

Acknowledgments

• Built for Void Editor
• Uses Model Context Protocol (MCP)
• Inspired by Anthropic's MCP examples

Support

• 📚 Void Editor Documentation
• 📖 MCP Documentation
• 🐛 Report Issues

Roadmap

• Backup/undo functionality
• Operation logging and audit trails
• Rate limiting
• Dry-run mode for previewing changes
• Git integration (auto-commit after changes)
• Multi-user support