AccuKnox Server

Complete guide for testing MCP servers with stdio client, Gemini CLI, HTTP client, and VSCode Copilot.

📋 Table of Contents

• Prerequisites
• Project Structure
• Installation
• Testing stdio Server
  • With Test Client
  • With Gemini CLI
  • With VSCode Copilot (stdio)
• Testing HTTP Server
  • With Test Client
  • With VSCode Copilot (HTTP)

---

Prerequisites

• Python 3.10 or higher
• pip package manager
• Gemini CLI (optional, for AI integration)
• VSCode or Cursor (optional, for IDE integration)

---

Project Structure

MCP_server/
├── MCP_server.py              # stdio server (for Gemini CLI)
├── MCP_server_http.py         # HTTP server (for web/remote)
├── clients/
│   ├── __init__.py
│   ├── stdio_client.py        # Test client for stdio
│   └── http_client.py         # Test client for HTTP
├── shared/
│   ├── __init__.py
│   ├── tools.py               # Tool implementations
│   └── api.py          # AccuKnox API client
├── .vscode/
│   └── mcp.json               # VSCode Copilot configuration
├── .env                       # Environment variables (create your own)
├── requirements.txt
└── README.md

---

Installation

1. Clone/Setup Project

cd MCP_server

2. Create Virtual Environment

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment

Create .env file:

ACCUKNOX_BASE_URL=https://cspm.demo.accuknox.com
ACCUKNOX_API_TOKEN=your_token_here

---

Testing stdio Server

The stdio server uses standard input/output for communication, designed for local CLI integration.

With Test Client

Terminal 1: Start server (automatically started by client)

Terminal 2: Run test client

python3 clients/stdio_client.py

Interactive Menu:

Options:
  1 - Count assets
  2 - List 5 assets
  3 - Search by category
  4 - Get vulnerabilities
  exit - Quit

Select: 1

Example Output:

Total assets: 1247

---

With Gemini CLI

Gemini CLI provides AI-powered natural language interactions with your MCP server.

1. Install Gemini CLI

pip install google-generativeai

2. Configure MCP Server

# Add server to Gemini CLI (use absolute path)
gemini mcp add accuknox python3 /absolute/path/to/MCP_server.py

3. Verify Configuration

gemini mcp list

# Output:
# accuknox: python3 /path/to/MCP_server.py

4. Start Gemini CLI

gemini

5. Test with Natural Language Queries

Example 1: Count Assets

You: How many cloud assets do I have?

Gemini: You have 1,247 cloud assets in your inventory.

Example 2: Search by Category

You: Show me 5 Container assets

Gemini: Here are 5 Container assets:
1. nginx-prod (Container) - Region: us-east-1
2. redis-cache (Container) - Region: us-west-2
3. postgres-db (Container) - Region: eu-west-1
4. kafka-broker (Container) - Region: us-east-1
5. redis-sentinel (Container) - Region: ap-south-1

Example 3: Security Vulnerabilities

You: What security vulnerabilities do my AI models have?

Gemini: Your AI models have 147 security issues:
• ML Models: 5 issues (1 Critical, 4 Medium)
• LLM Models: 136 issues (136 Critical)
• Datasets: 6 issues (6 High)

Immediate action required for 137 critical vulnerabilities.

6. Remove Server (Optional)

gemini mcp remove accuknox

---

With VSCode Copilot (stdio)

1. Create Configuration File

Create .vscode/mcp.json in your project root:

mkdir -p .vscode
nano .vscode/mcp.json

2. Add stdio Server Configuration

{
  "mcpServers": {
    "accuknox-stdio": {
      "command": "python3",
      "args": ["${workspaceFolder}/MCP_server.py"],
      "env": {
        "ACCUKNOX_BASE_URL": "https://cspm.demo.accuknox.com",
        "ACCUKNOX_API_TOKEN": "your_token_here"
      }
    }
  }
}

Note: Using ${workspaceFolder} makes the configuration portable across machines.

3. Activate in VSCode

1. Open your project in VSCode
2. VSCode automatically detects .vscode/mcp.json
3. Click "Start" when prompted
4. Open Copilot Chat: Ctrl+Shift+I (Windows/Linux) or Cmd+Shift+I (Mac)
5. Enable Agent Mode (click brain icon )
6. Click tools icon () → Enable accuknox-stdio

4. Test with Copilot

@agent How many cloud assets do I have?
@agent Show me Container assets in AWS
@agent What are my model vulnerabilities?

---

Testing HTTP Server

The HTTP server provides REST API access for web clients and remote integrations.

With HTTP Test Client

Terminal 1: Start HTTP server

python3 MCP_server_http.py

Output:

======================================================================
AccuKnox MCP Server - HTTP
======================================================================
Server: http://localhost:8000
Tools: search_assets, get_model_vulnerabilities
Press Ctrl+C to shutdown
======================================================================
INFO:     Uvicorn running on http://0.0.0.0:8000

Terminal 2: Run test client

python3 clients/http_client.py

Interactive Menu:

Options:
  1 - Count assets
  2 - List 5 assets
  3 - Search by category
  4 - Get vulnerabilities
  exit - Quit

Select: 4

---

With VSCode Copilot (HTTP)

VSCode Copilot supports HTTP MCP servers starting from version 1.102+.

Prerequisites

• VSCode 1.102 or higher
• GitHub Copilot extension installed
• HTTP MCP server running on http://localhost:8000

---

Step 1: Verify HTTP Server

# Start HTTP server
python3 MCP_server_http.py

# In another terminal, verify it's running
curl http://localhost:8000/health
# Expected: {"status": "healthy"}

---

Step 2: Configure VSCode

Option A: Project-Level Configuration (Recommended)

Create .vscode/mcp.json:

{
  "servers": {
    "accuknox-http": {
      "type": "http",
      "url": "http://localhost:8000"
    }
  },
  "inputs": []
}

Option B: Remote Server Configuration

For deployed servers:

{
  "servers": {
    "accuknox-http": {
      "type": "http",
      "url": "https://your-server.com:8000",
      "headers": {
        "Authorization": "Bearer ${input:api-token}"
      }
    }
  },
  "inputs": [
    {
      "type": "promptString",
      "id": "api-token",
      "description": "AccuKnox API Token",
      "password": true
    }
  ]
}

Option C: Global Configuration

For all workspaces, create ~/.vscode/mcp.json:

{
  "servers": {
    "accuknox-http": {
      "type": "http",
      "url": "http://localhost:8000"
    }
  },
  "inputs": []
}

---

Step 3: Activate Server

Method 1: Automatic Detection

1. Open project in VSCode
2. VSCode detects .vscode/mcp.json
3. Click "Start" button in the file
4. Server status shows "running" in status bar

Method 2: Command Palette

1. Press Ctrl+Shift+P (or Cmd+Shift+P on Mac)
2. Type: "MCP: Add Server"
3. Select "HTTP/SSE"
4. Enter Server ID: accuknox-http
5. Enter URL: http://localhost:8000
6. Choose Workspace or Global
7. Click Save

Method 3: Manual Configuration

1. Create .vscode/mcp.json with above content
2. Save file
3. VSCode auto-detects and shows Start button

---

Step 4: Enable Copilot Agent Mode

1. Open Copilot Chat:

   • Press Ctrl+Shift+I (Windows/Linux)
   • Press Cmd+Shift+I (Mac)
   • Or click Copilot icon in Activity Bar

2. Enable Agent Mode:

   • Click the brain icon () at top
   • Or type @agent in chat

3. Configure Tools:

   • Click tools icon () at bottom
   • Find accuknox-http in list
   • Toggle ON to enable

---

Step 5: Test with Copilot

Example 1: Asset Inventory

@agent How many cloud assets do I have?

Copilot: You have 1,247 cloud assets in your inventory.

Example 2: Category Search

@agent Show me 5 Container assets

Copilot: Here are 5 Container assets:
1. nginx-prod (Container) - Region: us-east-1
2. redis-cache (Container) - Region: us-west-2
3. postgres-db (Container) - Region: eu-west-1
4. kafka-broker (Container) - Region: us-east-1
5. redis-sentinel (Container) - Region: ap-south-1

Example 3: Security Analysis

@agent What security vulnerabilities do my AI models have?

Copilot: Your AI models have 147 security issues:
• ML Models: 5 issues (1 Critical, 4 Medium)
• LLM Models: 136 issues (136 Critical)
• Datasets: 6 issues (6 High)

Critical attention required for 137 vulnerabilities.

Example 4: Cloud Provider Filter

@agent List AWS assets in us-east-1 region

Copilot: Found 234 AWS assets in us-east-1 region...

---

Quick Reference

Server Commands

Action	Command
stdio server	python3 MCP_server.py
HTTP server	python3 MCP_server_http.py
stdio client	python3 clients/stdio_client.py
HTTP client	python3 clients/http_client.py

Configuration Paths

Tool	Config Location
VSCode Copilot	.vscode/mcp.json or ~/.vscode/mcp.json
Gemini CLI	Managed by gemini mcp commands