Windows Developer MCP Server

The production-grade, open-source Windows Developer Model Context Protocol (MCP) server.
Engineered to solve the dual challenge of Glama Quality Indexing and Local LLM Execution Efficiency.

---

🌟 Executive Overview

Windows Developer MCP is a secure, high-performance MCP server designed specifically for Windows development environments. It provides native integration with LM Studio, Claude Desktop, and other MCP-compatible clients.

Traditional MCP servers present a dilemma:

1. Directory Ranking (Glama) rewards exposing hundreds of fine-grained tools with extensive API coverage and comprehensive parameter schemas.
2. Local LLMs (Gemma, Qwen 2.5, DeepSeek R1, Llama 3) running in LM Studio have constrained context budgets (4K–16K tokens). Exposing 100+ raw tools saturates system prompts (~15K tokens), causing model paralysis, parameter hallucination, and high latency.

Windows Developer MCP solves this with a Dual-Profile Engine and Dynamic Capability Discovery:

• Glama Evaluators see a FULL API surface with hyper-structured Google/Sphinx style tool definitions.
• Local LLMs default to a MINIMAL profile (<10 core tools, <1.2K token footprint) and use dynamic meta-tools (mcp_search_tools, mcp_enable_provider) to discover and register domain capabilities on demand.

---

🏗️ System Architecture

flowchart TD
    Client["MCP Client (LM Studio / Claude Desktop)"]
    FastMCP["FastMCP Core Application (server.py)"]
    Registry["Provider Registry (core/registry.py)"]
    
    subgraph Profiles ["Execution Profiles"]
        P_Min["MINIMAL (Context Footprint: ~1.2K tokens)"]
        P_Dev["DEVELOPER (Balanced Footprint: ~3.8K tokens)"]
        P_Full["FULL (Glama Full Exposure: ~12K tokens)"]
    end
    
    subgraph Providers ["12 Single-Responsibility Domain Providers"]
        Terminal["TerminalProvider"]
        Filesystem["FilesystemProvider"]
        Git["GitProvider"]
        Docker["DockerProvider"]
        Python["PythonProvider"]
        Node["NodeProvider"]
        Windows["WindowsProvider"]
        SQLite["SQLiteProvider"]
        Network["NetworkProvider"]
        Project["ProjectProvider (AI Tools)"]
        GitHub["GitHubProvider"]
        Browser["BrowserProvider"]
    end

    subgraph Security ["Hardened Security Subsystem"]
        Sandbox["Workspace Sandbox (sandbox.py)"]
        Validator["Command Validator (validator.py)"]
        RateLimiter["Rate Limiter (rate_limiter.py)"]
        Logger["Audit Logger (logger.py)"]
    end

    Client -->|STDIO / JSON-RPC| FastMCP
    FastMCP --> Registry
    Registry --> Profiles
    Profiles --> Providers
    Providers --> Security
    Security -->|PowerShell 7 / Cmd Subprocess| WinOS["Windows OS"]

---

⚡ Dual-Profile Engine Comparison

Profile	Target Environment	Active Tools	Context Footprint	Description
minimal (Default)	Local LLMs (7B/8B) in LM Studio	~8 Core Tools + Meta-Tools	~1.2K Tokens	Low-overhead mode. Exposes core terminal, read/write, git, and project tools.
developer	Standard LLMs (14B/32B) / Claude Desktop	~25 Primary Dev Tools	~3.8K Tokens	Balanced toolset covering standard developer workflows.
full	70B+ LLMs / Glama Indexing Audits	All Registered Tools	~12.0K Tokens	Exposes exhaustive API surface across all 12 domain providers.

---

🔌 12 Domain Providers & Capabilities

Provider	Module	Description	Core Exposed Tools
Terminal	[providers/terminal.py](file:///c:/Users/Peeyush/terminal-mcp/providers/terminal.py)	Safe PowerShell/CMD subprocess execution	terminal_run, terminal_get_session
Filesystem	[providers/filesystem.py](file:///c:/Users/Peeyush/terminal-mcp/providers/filesystem.py)	Sandbox-restricted file reading, writing, tree view	filesystem_read, filesystem_write, filesystem_list
Git	[providers/git.py](file:///c:/Users/Peeyush/terminal-mcp/providers/git.py)	Working tree, commit log, diff, and branch tracking	git_status, git_log, git_diff
Python	[providers/python.py](file:///c:/Users/Peeyush/terminal-mcp/providers/python.py)	Virtual environment script execution and package check	python_run, python_check_package
Node.js	[providers/node.py](file:///c:/Users/Peeyush/terminal-mcp/providers/node.py)	Node execution and npm package management	node_run, npm_run
Docker	[providers/docker.py](file:///c:/Users/Peeyush/terminal-mcp/providers/docker.py)	Container, image, compose, and build orchestration	docker_list_containers, docker_logs
Windows	[providers/windows.py](file:///c:/Users/Peeyush/terminal-mcp/providers/windows.py)	System metrics, process list, and environment vars	windows_system_info, windows_list_processes
SQLite	[providers/sqlite.py](file:///c:/Users/Peeyush/terminal-mcp/providers/sqlite.py)	Safe SQLite database queries and schema inspection	sqlite_query, sqlite_schema
Network	[providers/network.py](file:///c:/Users/Peeyush/terminal-mcp/providers/network.py)	Diagnostic ping and DNS resolution	network_ping, network_dns_lookup
Project	[providers/project.py](file:///c:/Users/Peeyush/terminal-mcp/providers/project.py)	AI workspace analysis, security scan, docs & tests gen	project_analyze, project_security_scan
GitHub	[providers/github.py](file:///c:/Users/Peeyush/terminal-mcp/providers/github.py)	GitHub CLI wrapper for repos, PRs, issues	github_auth_status, github_repo_info
Browser	[providers/browser.py](file:///c:/Users/Peeyush/terminal-mcp/providers/browser.py)	Browser automation interface (disabled by default)	browser_navigate

---

🧠 Native AI Developer Tools

ProjectProvider exposes AI-native tools for codebase comprehension:

• 🔍 project_analyze: Detects frameworks, tech stacks, and key config files automatically.
• 📦 project_dependencies: Summarizes cross-language Python/Node dependencies.
• 🌳 project_summarize: Generates a hierarchical workspace file tree.
• 🏛️ project_architecture: Detects design patterns (Provider Pattern, Layered Architecture).
• 📝 project_generate_readme: Generates a complete, structured README markdown draft.
• 📚 project_generate_docs: AST-based docstring extraction for Python source files.
• 🛡️ project_security_scan: Static pattern audit for hardcoded API keys and credentials.
• 🧪 project_generate_tests: Generates pytest unit test skeletons for source files.

---

🛡️ Security & Sandbox Architecture

1. Workspace Boundary Sandbox ([security/sandbox.py](file:///c:/Users/Peeyush/terminal-mcp/security/sandbox.py)): Strict path normalization prevents path traversal out of workspace root (../..).
2. Command Validation Engine ([security/validator.py](file:///c:/Users/Peeyush/terminal-mcp/security/validator.py)): Regex denylists block destructive operations (shutdown, format c:, del /f /s /q, net user admin).
3. Rate Limiter ([security/rate_limiter.py](file:///c:/Users/Peeyush/terminal-mcp/security/rate_limiter.py)): Sliding-window execution limiter protects against command loops.
4. Structured Audit Logging ([security/logger.py](file:///c:/Users/Peeyush/terminal-mcp/security/logger.py)): All executed commands, timestamps, exit codes, and output sizes are logged to logs/audit.log.

---

🚀 Quick Start

1. Installation with uv

# Clone repository
git clone https://github.com/peeyushcodes/windows-developer-mcp.git
cd windows-developer-mcp

# Synchronize dependencies with uv
uv sync

2. Launch Server

uv run windows-developer-mcp

---

💻 LM Studio & Claude Desktop Integration

LM Studio Setup

Add to your LM Studio MCP server configuration file:

{
  "mcpServers": {
    "windows-developer-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\path\\to\\windows-developer-mcp",
        "run",
        "windows-developer-mcp"
      ],
      "env": {
        "MCP_SERVER__PROFILE": "minimal",
        "MCP_WORKSPACE__PATH": "C:\\path\\to\\workspace"
      }
    }
  }
}

Claude Desktop Setup

Add to %APPDATA%\Claude\claude_desktop_config.json:

{
  "mcpServers": {
    "windows-developer-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\path\\to\\windows-developer-mcp",
        "run",
        "windows-developer-mcp"
      ],
      "env": {
        "MCP_SERVER__PROFILE": "developer",
        "MCP_WORKSPACE__PATH": "."
      }
    }
  }
}

---

📖 Comprehensive Documentation Suite

• 📐 Architecture Overview
• 🔌 Provider Reference
• 🛡️ Security Model
• 🤝 Contributing Guide
• 💡 Usage Workflows & Examples
• 🤖 LM Studio Setup Guide
• 🖥️ Claude Desktop Setup Guide
• 🖼️ Screenshots & Visual Assets Plan
• 📋 Glama Manifest (glama.json)
• 📜 Changelog

---

📄 License

Distributed under the MIT License. See LICENSE for details.