universal-mcp-fs
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@universal-mcp-fs list files in my projects directory"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
universal-mcp-fs
A secure, local MCP server that gives AI assistants controlled access to your filesystem.
🛠️ Built With
Overview
universal-mcp-fs gives AI assistants — Claude Desktop today, and any other stdio-compatible MCP client as they add support — filesystem access and command execution on your machine, gated behind an interactive approval system.
stdio only. No HTTP server, no open ports, no internet exposure.
Elicitation-based approval. Dangerous actions (delete, run commands, move files) pause and ask the connected client to show a native approval popup — the same kind of Allow/Deny prompt you already see for other tool calls in Claude Desktop.
Sensitive paths are blocked outright (
.ssh,.aws,.gnupg, browser credential stores,/etc/shadow, etc.) — before any approval prompt is even offered.17 tools covering file read/write/move/copy/delete, directory listing, filename and content search, and shell command execution (foreground and background).
Related MCP server: Remote Server MCP
Install
Option 1 — npm (recommended, no build step)
npm install -g universal-mcp-fsOption 2 — from source (GitHub)
git clone https://github.com/Prabhas125/universal-mcp-fs.git
cd universal-mcp-fs
npm install
npm run buildSetup: Claude Desktop
Edit your Claude Desktop config file:
Windows:
%APPDATA%\Claude\claude_desktop_config.jsonLinux:
~/.config/Claude/claude_desktop_config.jsonmacOS:
~/Library/Application Support/Claude/claude_desktop_config.json
If you installed via npm (Option 1):
{
"mcpServers": {
"filesystem": {
"command": "universal-mcp-fs",
"args": ["--allowed-dirs", "/home/user;/home/user/projects"]
}
}
}If you built from source (Option 2):
{
"mcpServers": {
"filesystem": {
"command": "node",
"args": [
"/full/path/to/universal-mcp-fs/dist/index.js",
"--allowed-dirs",
"/home/user;/home/user/projects"
]
}
}
}On Windows, use double-escaped backslashes in the path, e.g. "C:\\Users\\yourname\\projects\\universal-mcp-fs\\dist\\index.js".
Then restart Claude Desktop. It will spawn the server automatically and the 17 tools will be available in chat. No further setup, no login, no token.
Verify it's connected
Ask Claude something like "list the files in [one of your allowed directories]". If it responds with a directory listing, the server is connected. If Claude says it has no matching tool, double check the config file path and JSON syntax, then fully quit and reopen Claude Desktop (a reload isn't enough — it only reads this file on startup).
Configuration options
All options can be passed as CLI args in the config's args array, or as environment variables in an env block.
CLI flag | Env var | Default | Description |
|
| your home directory | Semicolon-separated list of directories the server may access |
|
| commands enabled | Disables |
|
| delete enabled | Disables |
|
| 10485760 (10 MB) | Max bytes |
|
| 30000 | Default timeout for |
|
| 120000 | How long an approval popup waits before auto-denying, in ms |
|
| 50 | Cap on results from search tools |
Tool reference
Filesystem — read_file, read_file_lines, write_file, create_directory, list_directory, move_file, copy_file, delete_file, delete_directory
Search — search_files (by filename/glob), search_content (grep-like, with context lines)
Commands — run_command, run_command_background, list_processes, kill_process
Info — file_info, system_info
Full parameter docs are visible to the AI client automatically (and to you, via npx @modelcontextprotocol/inspector).
Permission system
Some tools always require your approval before running: delete_file, delete_directory, run_command, run_command_background, move_file, kill_process, and write_file when overwriting an existing file.
When one of these is called, the server sends an elicitation request to Claude Desktop, which renders it as a native approval dialog — the same UI you already see for regular tool-call confirmations. You can:
Approve — the action runs once
Approve + "always allow" — this exact action (same tool, same file/command) is silently approved from then on, persisted to
~/.universal-mcp-fs/always-allow.jsonDecline/cancel — the action is aborted, nothing happens
If you don't respond within 2 minutes, the request automatically resolves to denied — it will not hang the connection or leave Claude waiting indefinitely.
Sensitive paths (.ssh, .aws, .gnupg, browser credential files, /etc/shadow) are rejected before any approval prompt is offered — no amount of clicking "allow" gets past this list. You can extend it via config/default.json if you build from source.
Every decision (approved, declined, always-allow, timed out, blocked) is logged to ~/.universal-mcp-fs/audit.log and to stderr, visible in Claude Desktop's MCP server logs.
Security model
No network transport — this server never opens a port or listens for external connections.
All file operations are validated against
--allowed-dirs; anything outside is rejected regardless of approval.Symlinks are resolved before validation, so a symlink inside an allowed directory can't be used to escape it.
Path traversal (
../..) is normalized away before any check runs.If the connected client doesn't support elicitation, dangerous tools fail closed (denied), never fail open.
Publishing / contributing
npm run build
npm version patch # bumps package.json, creates a git tag
git push && git push --tags
npm publishIssues and PRs welcome.
License
MIT
Available Tools
17 toolscopy_fileCopy File or DirectoryA
Copy a file or directory (recursively for directories). Does not modify the source, so no approval is required.
Args:
source (string): Path to copy from.
destination (string): Path to copy to.
Inputs: source (required), destination (required).
create_directoryCreate DirectoryAIdempotent
Create a directory, including any missing parent directories (like mkdir -p).
Args:
path (string): Directory path to create.
Returns: confirmation of the created path. No error if the directory already exists.
Inputs: path (required).
delete_directoryDelete DirectoryADestructive
Permanently delete a directory and all its contents, recursively. ALWAYS requires interactive user approval before proceeding — this cannot be undone.
Args:
path (string): Directory to delete.
Inputs: path (required).
delete_fileDelete FileADestructive
Permanently delete a single file. ALWAYS requires interactive user approval before proceeding — this cannot be undone.
Args:
path (string): Path to the file to delete.
Inputs: path (required).
file_infoGet File InfoARead-onlyIdempotent
Get metadata about a file or directory: size, created/modified dates, permissions, and type.
Args:
path (string): Path to inspect.
Returns: size in bytes, created/modified timestamps, POSIX permission bits, and whether it's a directory or symlink.
Inputs: path (required).
kill_processKill Background ProcessADestructive
Terminate a background process previously started with run_command_background. Requires interactive user approval before proceeding.
Args:
pid (number): Process ID to kill (must be one tracked by this server — see list_processes).
Inputs: pid (required).
list_directoryList DirectoryARead-onlyIdempotent
List the contents of a directory with metadata (type, size, modified date).
Args:
path (string): Directory to list.
recursive (boolean, optional): List subdirectories recursively. Default: false.
show_hidden (boolean, optional): Include dotfiles/dotdirs. Default: false.
Returns: a listing with one entry per line: type, size, modified date, name.
Inputs: path (required), recursive, show_hidden.
list_processesList Background ProcessesARead-onlyIdempotent
List background processes started via run_command_background during this session, with their status.
Returns: a list of tracked processes with PID, command, status, and start time.
Inputs: no named parameters.
move_fileMove or Rename FileADestructive
Move or rename a file or directory. ALWAYS requires interactive user approval before proceeding.
Args:
source (string): Current path.
destination (string): Target path.
Error Handling:
Both source and destination are validated against allowed directories and blocked paths.
Returns an error if the user declines the approval prompt.
Inputs: source (required), destination (required).
read_fileRead FileARead-onlyIdempotent
Read the full contents of a file. Text files are returned as UTF-8 text; binary files are detected automatically and returned as base64.
Args:
path (string): Absolute or relative path to the file.
Returns: file contents as text, or a note that the file is binary with base64 content.
Error Handling:
Returns an error if the path is outside allowed directories or matches a blocked pattern (e.g. .ssh, .aws).
Returns "File or directory not found" if the path does not exist.
Inputs: path (required).
read_file_linesRead File LinesARead-onlyIdempotent
Read a specific line range from a file. Useful for large files where reading the whole thing would exceed limits.
Args:
path (string): Path to the file.
start_line (number): 1-indexed starting line (inclusive).
end_line (number): 1-indexed ending line (inclusive).
Returns: the requested lines, each prefixed with its line number.
Inputs: path (required), end_line (required), start_line (required).
run_commandRun Shell CommandADestructive
Execute a shell command and return its stdout/stderr. ALWAYS requires interactive user approval before running.
Args:
command (string): The command to execute (interpreted by the platform shell — cmd.exe on Windows, bash on Linux).
cwd (string, optional): Working directory. Defaults to the first allowed directory.
timeout (number, optional): Timeout in milliseconds. Default: 30000, max 300000.
Returns: stdout, stderr, and exit code.
Error Handling:
Returns an error if the user declines the approval prompt.
Returns an error if the command exceeds the timeout (process is killed).
Inputs: command (required), cwd, timeout.
run_command_backgroundRun Command in BackgroundADestructive
Start a long-running command in the background and return immediately with its process ID. ALWAYS requires interactive user approval before running.
Args:
command (string): The command to execute.
cwd (string, optional): Working directory. Defaults to the first allowed directory.
Returns: the process ID (pid), which can be used with list_processes and kill_process.
Note: background processes have no timeout, but they are tracked and can be killed with kill_process. They do not survive this server process restarting (e.g. Claude Desktop being closed and reopened).
Inputs: command (required), cwd.
search_contentSearch File ContentsARead-onlyIdempotent
Grep-like search: find a text query across files in a directory, returning matching lines with 2 lines of context before/after.
Args:
directory (string): Base directory to search within.
query (string): Text to search for (case-sensitive, literal substring match).
file_pattern (string, optional): Glob to limit which files are searched (e.g. "**/*.js"). Default: all files.
max_results (number, optional): Cap on returned matches. Default: 50.
Returns: matching lines with file path, line number, and surrounding context.
Inputs: query (required), directory (required), max_results, file_pattern.
search_filesSearch Files by NameARead-onlyIdempotent
Find files matching a glob pattern within a directory (e.g. "**/.ts", ".log").
Args:
directory (string): Base directory to search within.
pattern (string): Glob pattern to match filenames against.
max_results (number, optional): Cap on returned results. Default: 50.
Returns: a list of matching file paths, relative to the search directory.
Inputs: pattern (required), directory (required), max_results.
system_infoGet System InfoARead-onlyIdempotent
Get information about the machine this server is running on: OS, hostname, home directory, Node version, CPU count, and total memory.
Returns: a JSON object with platform details.
Inputs: no named parameters.
write_fileWrite FileADestructive
Create a new file or overwrite an existing one. Parent directories are created automatically.
Overwriting an EXISTING file requires interactive user approval (elicitation) — the server will pause and ask the connected client to confirm before proceeding.
Args:
path (string): Path to write to.
content (string): Text content to write (UTF-8).
overwrite (boolean, optional): Must be true to overwrite an existing file. Default: false.
Error Handling:
Returns an error without writing if the file exists and overwrite is not true.
Returns an error if the user declines the overwrite approval prompt.
Inputs: path (required), content (required), overwrite.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 17 tools
Each tool has a clearly distinct purpose: read_file vs read_file_lines differ by scope; write_file, move_file, copy_file, delete_file, delete_directory are all unique operations; search_files vs search_content are different search types; run_command vs run_command_background vs process management tools are separate; system_info and file_info are metadata distinct from content tools. No overlap.
All tool names follow a consistent verb_noun pattern in snake_case (e.g., read_file, list_directory, search_content, kill_process). No mixing of camelCase or other conventions, making the namespace predictable.
17 tools is slightly above the typical 3-15 range, but each tool covers a distinct operation (file read/write, directory ops, searching, command execution, process management, system info) and the scope justifies the count. It feels well-scoped for a file system server.
The tool set covers all essential file system operations: create, read, update (overwrite), delete for files; directory creation and deletion; move/copy; file and content search; metadata retrieval; and even command execution with process management. No obvious gaps for common tasks.
Related MCP Connectors
Remote MCP server exposing your Node Command fleet (files, shell, processes, browser) to AI hosts.
MCP server that lets your AI assistant build and deploy apps, hosting each one at a live URL.
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to interact with yo…
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceAn enhanced MCP server that grants AI assistants the ability to execute terminal commands on a user's system with improved security controls, designed for use in controlled environments.2MIT
- AlicenseNot gradedqualityFmaintenanceA security-first MCP server that gives AI assistants controlled, safe access to manage remote servers via SSH with whitelisted operations and no generic command execution.MIT
- AlicenseAqualityCmaintenanceAn MCP server that lets an AI assistant run commands and transfer files over SSH without exposing connection details.5MIT
- FlicenseBqualityCmaintenanceA lightweight MCP server that enables AI assistants to interact with the local machine through terminal, filesystem, and Python execution tools.9-