enhanced-filesystem-mcp
Provides Git operations including status, diff, log, blame, and commit.
Manages NVIDIA GPU stats and services.
Integrates with Ollama for LLM generation, chat, model pulling, and management.
Runs Python code with persistent sessions.
Manages TOML configuration files (part of YAML/TOML tool).
Analyzes TypeScript/JS code structure.
Manages YAML configuration files (part of YAML/TOML tool).
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., "@enhanced-filesystem-mcp show me the git status of my project"
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.
Enhanced Filesystem MCP Server
High-performance Model Context Protocol (MCP) server giving AI agents advanced filesystem and automation capabilities on Windows.
26 tools across file I/O, search, Git, process management, Ollama, HTTP, SSH, Windows UI automation, and more — all from a single server.
Install
Requirements: Node.js >= 22 · Windows (for windows_tool; other tools are cross-platform)
git clone https://github.com/AesirKode/enhanced-filesystem-mcp
cd enhanced-filesystem-mcp
npm install
npm run buildConnect to Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"enhanced-filesystem": {
"command": "node",
"args": ["C:/path/to/enhanced-filesystem-mcp/dist/index.js"],
"env": {
"ALLOWED_DIRS": "C:/Users/you/Projects",
"MAX_PARALLEL": "20"
}
}
}
}Related MCP server: Windows-MCP
The Toolkit (26 Tools)
File Operations
Tool | Purpose |
| Read any file with caching and pagination |
| Write files of any size — no chunking |
| Surgical find/replace |
| Directory listing with depth and pattern filters |
| File metadata: size, type, timestamps |
| Delete files or directories |
| Move or rename |
| Parallel file/content search |
| Atomic multi-operation execution with rollback |
| Git: status, diff, log, blame, commit |
Execution & Process
Tool | Purpose |
| Shell command execution |
| Python with persistent sessions |
| Service/GPU management + NVIDIA stats |
AI & Integration
Tool | Purpose |
| Ollama LLM: generate, chat, pull, manage models |
| HTTP client with auth, retry, file download |
| Deep JSON with JSONPath queries |
Utilities
Tool | Purpose |
| Zip: create, extract, list, add |
| MD5 / SHA1 / SHA256 / SHA512 |
| System clipboard read/write |
| Smart downloads from CivitAI, HuggingFace, URLs |
| GGUF/Safetensors inspection without loading |
| YAML/TOML config management |
| File and directory comparison |
| Windows UI automation: keyboard, mouse, screenshots |
| TypeScript/JS code structure analysis |
| SSH client with SFTP |
Configuration
All settings via environment variables:
Variable | Default | Description |
| (none) | Comma-separated allowed directories |
|
| LRU cache entries |
|
| Max parallel operations |
|
| Base path for |
Contributing
See CONTRIBUTING.md. Bug reports and feature requests welcome via GitHub Issues.
License
MIT — Aesir / David R.
Available Tools
26 toolsanalysis_toolA
Code analysis tool for TypeScript/JavaScript files.
Extracts structure (classes, functions, imports) without reading the full file. Useful for understanding codebases quickly.
Operations:
'summarize': Get high-level structure of a file
'imports': List all imports in a file
Examples:
Summarize a file: { operation: 'summarize', path: 'src/index.ts' }
Check imports: { operation: 'imports', path: 'src/core/utils.ts' }
Inputs: path (required), operation (required).
archive_toolA
Zip archive operations: list contents, create from files/dirs, extract to directory, add files to existing archive.
Inputs: operation (required), path, files, output, +2 more.
clipboard_toolB
System clipboard operations: copy text, read/paste current content, check type, clear, copy file paths for Explorer, get file paths from clipboard.
Inputs: operation (required), text, files.
diff_toolA
Compare files and directories. Operations: files (line-by-line diff), dirs (structural comparison), stat (size/date metadata), quick (fast identical check). Supports unified diff format.
Inputs: path1 (required), path2 (required), operation (required), limit, +2 more.
download_toolA
Smart model downloads with resume, progress, SHA256 verification, and auto-placement. Sources: CivitAI (auto-detects type), HuggingFace, direct URLs. Operations: download, info (preview without downloading), list-paths (show all model folders).
Inputs: operation (required), url, hash, filename, +1 more.
efs_batchA
Execute multiple file operations atomically.
Features:
All operations succeed or all rollback
Automatic backups before modifications
Parallel execution when possible
Transaction support
All operations complete in a single call with automatic error handling.
Operations supported:
read: Read file
write: Create/modify file
edit: Find/replace in file
delete: Remove file
move: Move/rename file
list: List directory
Parameters:
operations: Array of operations to execute
atomic: Use transactions (all-or-nothing) (default: true)
parallel: Execute in parallel when safe (default: true)
Operation format: { op: 'read' | 'write' | 'edit' | 'delete' | 'move' | 'list', path: string, // op-specific fields... }
Examples:
Read multiple files: { operations: [ { op: 'read', path: 'file1.txt' }, { op: 'read', path: 'file2.txt' }, { op: 'read', path: 'file3.txt' } ], parallel: true }
Atomic refactoring: { operations: [ { op: 'edit', path: 'file1.js', oldText: 'oldName', newText: 'newName' }, { op: 'edit', path: 'file2.js', oldText: 'oldName', newText: 'newName' }, { op: 'move', source: 'old.js', destination: 'new.js' } ], atomic: true }
Project setup: { operations: [ { op: 'write', path: 'src/index.ts', content: '...' }, { op: 'write', path: 'package.json', content: '...' }, { op: 'write', path: 'tsconfig.json', content: '...' } ] }
Inputs: operations (required), atomic, parallel.
efs_deleteA
Delete file or directory. Set recursive: true for directories. Creates automatic backup before deletion.
Inputs: path (required), recursive.
efs_editA
Surgical find/replace in a file. oldText must be unique. count: default 1, use -1 for all occurrences. Set dryRun=true to preview the change without writing. Creates automatic backup.
Inputs: path (required), newText (required), oldText (required), count, +2 more.
efs_execB
Execute a shell command and return stdout/stderr. Parameters: command (required), cwd (optional), timeout ms (default 30000).
Inputs: command (required), cwd, timeout.
efs_gitB
Git operations: status, diff, log, blame, add, commit, branch, stash. repoPath required for all operations.
Inputs: repoPath (required), operation (required), args, count, +3 more.
efs_infoA
Get file/directory metadata: size, type, line count, created/modified timestamps.
Inputs: path (required).
efs_listA
List directory contents. Supports depth control, glob pattern filter, and sort by name/size/date.
Inputs: path (required), depth, sortBy, pattern.
efs_moveA
Move or rename a file or directory. Works across directories. Creates automatic backup.
Inputs: source (required), destination (required).
efs_pythonA
Execute Python code. Supports persistent named sessions (variables kept between calls). Parameters: code (required), sessionId, cwd, timeout, restart.
Inputs: code (required), cwd, restart, timeout, +1 more.
efs_readA
Read file with caching and pagination. Supports any size, text/binary, offset/length. Excel: specify sheet/range. PDF: extracts as markdown.
Inputs: path (required), range, sheet, length, +2 more.
efs_searchA
Fast parallel search with indexing.
Features:
Parallel search across all CPU cores (20x faster on your i7-14700K)
Smart indexing for frequently searched directories
Streaming results (see matches as they're found)
File name and content search
Regex and literal matching
Fuzzy search option
Search Types:
'files': Search by filename/pattern
'content': Search text inside files
'smart': Auto-detect best search type
Parameters:
path: Directory to search
pattern: Search pattern (regex or literal)
searchType: 'files' | 'content' | 'smart' (default: smart)
filePattern: Filter by file type (e.g., "*.js")
maxResults: Limit results (default: 100)
contextLines: Lines of context for content search (default: 2)
caseSensitive: Case sensitive search (default: false)
Examples: efs_search({ path: "D:/Projects", pattern: ".py", searchType: "files" }) efs_search({ path: "D:/Projects", pattern: "function_name", searchType: "content" }) efs_search({ path: "D:/Projects", pattern: "bug", filePattern: ".js", maxResults: 50 })
Inputs: path (required), pattern (required), maxResults, searchType, +3 more.
efs_writeA
Write file of any size — no chunking required. Mode: 'rewrite' (default) or 'append'. Invalidates cache automatically.
Inputs: path (required), content (required), mode, encoding.
hash_toolA
File hashing and verification. Operations: hash (single file), verify (against expected), compare (two files), multiple (batch), string (hash text). Algorithms: md5, sha1, sha256, sha512.
Inputs: operation (required), data, path, file1, +4 more.
http_toolA
HTTP client: GET/POST/PUT/DELETE/PATCH/HEAD. Supports JSON body, Bearer/Basic auth, custom headers, timeout, retry with backoff, file download via 'download' path param.
Inputs: url (required), method (required), auth, body, +4 more.
json_toolA
Deep JSON operations with JSONPath ($.key, $..key, $.arr[*], filter expressions). Operations: get, set, delete, merge (deep/shallow/replace), diff, flatten, unflatten, validate.
Inputs: path (required), operation (required), query, value, +6 more.
model_toolA
Inspect AI model files (Safetensors/GGUF) without loading into memory. Operations: info (metadata, tensors, quantization), list (scan directory), compare (two models), search (by name/arch).
Inputs: operation (required), path, limit, path1, +5 more.
ollama_toolA
Ollama LLM integration. Operations: list (installed models), show (model details), pull (download), delete, copy, create (from Modelfile), generate (single prompt), chat (multi-turn), embeddings, ps (loaded in memory). Config: optional host/timeout overrides.
Inputs: operation (required), model, config, prompt, +6 more.
process_toolA
Process and service manager with GPU monitoring. Operations: list (filter by name/port), gpu (NVIDIA stats), start (service or custom command), kill (by PID/name/port), status, restart, monitor, wait. Predefined services: ollama (11434), comfyui (8188), koboldcpp (5001).
Inputs: operation (required), cwd, env, pid, +7 more.
ssh_toolB
SSH client with persistent sessions. Operations: connect (host/username/password or privateKeyPath), exec (run command), upload/download (SFTP), list (remote dir), disconnect. sessionId required for all ops.
Inputs: operation (required), sessionId (required), host, port, +7 more.
windows_toolB
Windows UI automation. Keyboard: type, key (with modifiers: ctrl/alt/shift/win), hotkey (e.g. 'ctrl+c'). Mouse: click (left/right/middle, x/y coords), move, scroll. Windows: list, focus, minimize, maximize, close, active, window_info, move_window, screen_size. Other: screenshot (full or region), launch app.
Inputs: operation (required), x, y, key, +16 more.
yaml_toolB
YAML/TOML/JSON config file operations. Supports get (dot-notation query), set, delete, validate, convert between formats, merge, diff, keys, format. Auto-detects format from extension.
Inputs: operation (required), path, limit, query, +9 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 26 tools
Most tools have clearly distinct purposes, but efs_batch duplicates operations in efs_read/write/edit/delete/move/list, creating ambiguity about when to use batch versus individual tools. JSON and YAML tools overlap on JSON handling, and efs_exec/process_tool have unclear boundaries for shell versus process management.
The efs_ prefix is used for filesystem operations but also for git, exec, and python tools, making the prefix meaningless. Other tools use a '_tool' suffix with varied noun names, mixing snake_case conventions and creating two completely different naming patterns.
26 tools is excessive for a filesystem-focused server, especially since many tools (ollama, http, windows automation, ssh) are unrelated to files. The broad scope makes the set feel like a grab-bag rather than a cohesive toolset.
Core filesystem operations like read, write, edit, delete, move, list, and search are covered, but there is no copy tool, no chmod or symlink support, and efs_git includes only a subset of git commands. The non-filesystem tools feel bolted on and do not fill these primary domain gaps.
Maintenance
Related MCP Connectors
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
214 tools for AI agents over one MCP endpoint: image, video, audio, 3D, data, storage, mail, GPUs.
One MCP server for the services your AI needs: memory, files, Microsoft 365, kanban and more.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceA lightweight MCP HTTP server giving AI assistants real tools to interact with your Windows machine, including running commands, file access, system info, web search, and browser automation.1-
- AlicenseNot gradedqualityDmaintenanceA lightweight MCP server that enables AI agents to seamlessly integrate with Windows, performing tasks like file navigation, application control, UI interaction, and QA testing.1MIT
- FlicenseNot gradedqualityDmaintenanceA powerful filesystem MCP server for AI agents with extensive system access, including filesystem operations, shell execution, Windows tools, reverse engineering, code intelligence, and agent orchestration.1-
- AlicenseNot gradedqualityBmaintenanceEnables AI clients to securely control and interact with a local Windows machine through 218 configurable tools for files, Git, processes, Windows UI, browser automation, WSL, Office, recovery, skills, and child MCP servers.11 npmMIT