Give an AI assistant scoped read/write access to specific folders on your machine.
Preview full content
Lets an AI assistant read and write files within specific directories you allow — useful for having an assistant work directly with local project files, notes, or documents instead of copy-pasting content back and forth.
Add this to your MCP client's config (e.g. Claude Desktop's claude_desktop_config.json):
```json
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-filesystem",
"/absolute/path/to/allowed/folder"
]
}
}
}
```
Only the folder(s) you list are accessible — the assistant can't read or write outside them. List multiple folders as additional arguments if you need more than one. Restart your MCP client after editing the config for it to take effect.
Package names and exact flags shift as the MCP ecosystem evolves — treat this as the shape of the config (the mcpServers block, command/args/env keys), and check the server's current README before relying on it, especially the exact package name and version.