Filesystem MCP
Official Anthropic server for secure filesystem access — read, write, search, and manage local files from any MCP-compatible AI.
// Add to your client
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/Documents", "/Users/you/Desktop"]
}
}
}Paste into your client's MCP configuration file.
// Try it now
"List all files in my current project folder and tell me what kind of project it is"
- ! The server only sees directories you explicitly whitelist — paths outside are blocked
- ! Write access is powerful — AI can overwrite files, so scope to specific folders
- ! Symlinks can escape whitelisted folders — follow the server's path-validation guidance
No configuration needed
Completely free and open source. No API keys required.
// When to use
- You want Claude to read and understand your local project files
- You need AI to generate and save files to your computer
- You want to search through large numbers of local files
- You're building automations that work with local data
// When NOT to use
- You need to access files on a remote server (use SSH instead)
- You're working with sensitive data without reviewing permission boundaries
- You want cloud storage access (use Google Drive or Notion MCPs)
// Usage Scenarios
Project Analysis
Ask Claude to read your entire codebase, understand its structure, and give you a summary.
Example prompt
"Read all TypeScript files in ~/projects/myapp and tell me the architecture"
Content Generation
Generate documents, reports, or code files and save them directly to your chosen folder.
Example prompt
"Create a weekly report based on my notes in ~/notes/2025-02/ and save it as report.md"
// About
Plain English
This server gives your AI assistant the ability to see and edit files on your computer — like giving it access to your file manager. You decide which folders it can access. Then you can ask it to 'summarize all notes in my Documents folder' or 'create a new file with this content' and it does it directly on your machine.
The official Anthropic Filesystem MCP server provides Claude and other AI assistants with controlled access to your local filesystem. Configure allowed directories, and the AI can read files, write content, list directories, search for patterns, and move files — all within security boundaries you define.
This is one of the most widely-used MCP servers and the canonical example of how MCP works in practice.
// Use Cases
- Read and analyze local codebases with Claude
- Let AI write generated content directly to disk
- Search project files for patterns without leaving Claude
- Automate file organization and renaming tasks