Skip to content

README.md

Latest commit

 

History

History
177 lines (130 loc) · 4.73 KB

README.md

File metadata and controls

177 lines (130 loc) · 4.73 KB

CytoScnPy MCP Server

MCP (Model Context Protocol) server that exposes CytoScnPy's Python static analysis capabilities to LLMs like Claude, GPT, GitHub Copilot, and other AI assistants.

What is MCP?

The Model Context Protocol is an open standard that allows AI assistants to use external tools. This server enables LLMs to analyze Python code for:

  • Unused code - Functions, classes, imports, variables
  • Secrets - Hardcoded API keys, passwords, tokens
  • Dangerous patterns - eval, exec, SQL injection
  • Code quality - Cyclomatic complexity, maintainability index

Installation

Option 1: Standalone CLI (Recommended)

The MCP server is available in the standalone CLI binary:

# Install (Linux/macOS)
curl -fsSL https://raw.githubusercontent.com/djinn-soul/CytoScnPy/main/install.sh | bash

# Run MCP server (standalone CLI)
cytoscnpy mcp-server

The Python cytoscnpy package does not run mcp-server.

Option 2: VS Code Extension (Automatic)

Install the CytoScnPy VS Code extension. It automatically registers the MCP server with GitHub Copilot—no manual configuration required.

Option 3: Install Script (Standalone Binary)

Linux / macOS:

# Install
curl -fsSL https://raw.githubusercontent.com/djinn-soul/CytoScnPy/main/install.sh | bash

# Start MCP server
cytoscnpy mcp-server

Windows (PowerShell):

# Install
irm https://raw.githubusercontent.com/djinn-soul/CytoScnPy/main/install.ps1 | iex

# Start MCP server (after restarting terminal)
cytoscnpy mcp-server

Build from Source

cargo build --release -p cytoscnpy-cli

# Run MCP server
./target/release/cytoscnpy-cli mcp-server

Usage

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "cytoscnpy": {
      "command": "cytoscnpy",
      "args": ["mcp-server"]
    }
  }
}

Cursor IDE

Add to Cursor's MCP settings:

{
  "cytoscnpy": {
    "command": "cytoscnpy",
    "args": ["mcp-server"]
  }
}

GitHub Copilot (VS Code)

The VS Code extension automatically registers the MCP server. Just install the extension and ask Copilot:

"Run a quick security scan on this file using CytoScnPy"

Available Tools

Tool Description Parameters
analyze_path Full analysis on files/directories path, scan_secrets, scan_danger, check_quality
analyze_code Analyze code snippet directly code, filename
quick_scan Fast security scan (secrets & dangerous patterns) path
cyclomatic_complexity Calculate complexity metrics path
maintainability_index Calculate MI scores (0-100) path

Example Tool Calls

Quick security scan:

{
  "tool": "quick_scan",
  "arguments": {
    "path": "/home/user/myproject"
  }
}

Full analysis:

{
  "tool": "analyze_path",
  "arguments": {
    "path": "/home/user/myproject",
    "scan_secrets": true,
    "scan_danger": true,
    "check_quality": true
  }
}

Analyze code snippet:

{
  "tool": "analyze_code",
  "arguments": {
    "code": "def unused():\n    pass\n\ndef main():\n    print('hello')\n",
    "filename": "example.py"
  }
}

Architecture

┌─────────────────────┐
│  Claude / Copilot   │
│  (MCP Client)       │
└──────────┬──────────┘
           │ JSON-RPC over stdio
           ▼
┌─────────────────────┐
│  cytoscnpy          │
│  mcp-server         │
└──────────┬──────────┘
           │ Direct Rust function calls
           ▼
┌─────────────────────┐
│  cytoscnpy library  │
│  (Core analysis)    │
└─────────────────────┘

Future: HTTP Transport

HTTP/SSE transport is planned for remote LLM integrations. See roadmap.md for details.

License

Apache-2.0