网络MCP Docker套件

📋 Prerequisites

• Docker Engine 20.10+
• Docker Compose 2.0+
• API Access: Valid credentials for the network platforms you want to integrate (see individual server guides for specific requirements)

🌐 Network MCP Docker Suite

📚 Example Code for Learning & Development This is a demonstration project showcasing MCP server implementations for network management. Intended for educational purposes, testing, and development environments.

Docker-based MCP server suite for AIOps - enabling AI-driven network operations through Cisco Meraki, Catalyst Center, IOS XE, ISE, ThousandEyes, Splunk, NetBox & GitLab integration. AI-ready with LibreChat, Cursor, and other MCP clients for intelligent network management, automated troubleshooting, CI/CD orchestration, and operational insights.

📐 Deployment Architecture

The suite provides direct access to seven containerized MCP servers, perfect for development, testing, and AI-powered network operations:

┌─────────────────┐    ┌──────────────────────────────────┐
│                 │    │          Docker Host             │
│   MCP Client    │    │                                  │
│                 │    │  ┌─────────────────────────────┐ │
│ • Cursor IDE    │────┼─▶│ Meraki MCP        :8000     │ │
│ • LibreChat     │    │  ├─────────────────────────────┤ │
│ • Claude Desktop│────┼─▶│ NetBox MCP        :8001     │ │
│ • Other MCP     │    │  ├─────────────────────────────┤ │
│   Clients       │────┼─▶│ Catalyst Center   :8002     │ │
│                 │    │  ├─────────────────────────────┤ │
│                 │────┼─▶│ IOS XE MCP        :8003     │ │
│                 │    │  ├─────────────────────────────┤ │
│                 │────┼─▶│ ThousandEyes MCP  :8004     │ │
│                 │    │  ├─────────────────────────────┤ │
│                 │────┼─▶│ ISE MCP           :8005     │ │
│                 │    │  ├─────────────────────────────┤ │
│                 │────┼─▶│ Splunk MCP        :8006     │ │
│                 │    │  ├─────────────────────────────┤ │
│                 │────┼─▶│ Prometheus MCP    :8007     │ │
│                 │    │  ├─────────────────────────────┤ │
│                 │────┼─▶│ ClickHouse MCP    :8008     │ │
│                 │    │  ├─────────────────────────────┤ │
│                 │────┼─▶│ GitLab MCP        :8009     │ │
│                 │    │  └─────────────────────────────┘ │
└─────────────────┘    └──────────────────────────────────┘
        
        Direct HTTP Connections
        ✅ Simple setup - no authentication required
        ✅ Individual server access and configuration
        ✅ Flexible port-based deployment
        ✅ Perfect for development and testing

⚡ 3-Minute Setup

```bash

3. Deploy servers

./deploy.sh start all # All servers

4. Verify deployment

curl http://localhost:8000/mcp # Test Meraki server curl http://localhost:8002/mcp # Test Catalyst Center server ```

💡 Quick Tip: All servers now use a single centralized .env file for configuration. Use ENABLE_*_MCP=false to disable servers you don't need, and only add credentials for enabled servers.

🌐 LibreChat Integration: To use with LibreChat on an external network, see the External Network Integration section below.

🎯 Deployment Options

Deployment Examples

```bash

Flexible deployment using profiles

./deploy.sh start all # Complete suite ./deploy.sh start cisco # Cisco platforms only ./deploy.sh start monitoring # Monitoring focus ./deploy.sh start security # Security focus

3. Deploy (automatically uses override file)

./deploy.sh start all ```

The docker-compose.override.yml configures all MCP servers to join the external mcp-server network, allowing seamless communication with LibreChat and other services on the same network.

Deploy services

./deploy.sh start all # All servers ./deploy.sh start cisco # Cisco platforms ./deploy.sh start monitoring # Monitoring focused

Update and rebuild

git pull # Get updates docker-compose up -d --build # Rebuild and restart ```

Rebuild and restart

docker-compose up -d --build

🎯 Use Case

Network administrators and DevOps teams face significant challenges in managing modern hybrid network infrastructure across cloud and on-premises environments. This solution addresses these challenges by providing:

🚀 Primary Use Cases

#### 1. Unified Network Operations 🌐 - Single Interface: Manage Meraki cloud networks, on-premises NetBox DCIM/IPAM, Catalyst Center infrastructure, and direct IOS-XE devices through one MCP protocol interface - Streamlined Workflows: Reduce context switching between multiple network management tools and dashboards - Cross-Platform Visibility: Correlate data across different network management systems for comprehensive operational insights

#### 2. AI-Powered Network Management 🤖 - Natural Language Queries: Use AI assistants (Cursor, LibreChat) to query network infrastructure using plain English - Automated Troubleshooting: Enable AI-driven network issue diagnosis by providing unified access to network data - Intelligent Documentation: Generate automated reports combining real-time network state with infrastructure documentation

#### 3. DevOps Integration & Automation ⚙️ - Infrastructure as Code: Programmatic access to network infrastructure for automation workflows - CI/CD Integration: Embed network management capabilities into deployment pipelines - Configuration Management: Standardized API access for network device configuration and monitoring

#### 4. Operational Efficiency 📈 - Role-Based Access: Granular permissions for NOC teams (monitoring + firmware), SysAdmins (read-only), and full API access - Audit Trail: Comprehensive logging of all network management operations for compliance - Real-Time Synchronization: Automated synchronization between network devices and documentation systems

🚀 Quick Start

See .env.example for detailed configuration instructions

💻 Usage

🤖 Example Prompts

Here's a real-world example of how to interact with the MCP servers using natural language:

Network Troubleshooting Example

User Prompt:

Check why wlsn-access-1.dna.its-best.ch is unreachable from Cisco Catalyst Center.

AI Assistant Response: The AI assistant automatically uses both MCP servers working together:

1. Catalyst Center MCP Server - Checks device status and issues
2. IOS XE MCP Server - Direct SSH access to verify physical layer
3. Multi-Server Correlation - AI correlates data to identify root cause

Resolution Identified: - ✅ Device is UP and operational (verified via SSH) - ✅ Physical connectivity confirmed via CDP - ❌ Root Cause: IP address mismatch in Catalyst Center inventory - 🔧 Solution: Update device IP and re-sync

More Example Prompts

Monitor resource usage

docker stats

🎬 Live Demo

AI-Powered Network Troubleshooting with LibreChat using Multiple MCP Servers

Watch how natural language queries automatically investigate and resolve network issues using both Catalyst Center MCP Server and IOS XE MCP Server. The AI assistant correlates data from management systems (Catalyst Center) with direct device access (IOS XE SSH) to identify root causes and provide comprehensive solutions.

2. Configure environment variables (single .env file for all servers)

cp .env.example .env # Copy the environment template nano .env # Edit and configure: # - Set ENABLE_*_MCP=false for servers you don't want to use # - Add API keys and credentials for enabled servers

Enable/Disable individual servers (edit .env)

ENABLE_MERAKI_MCP=true # Set to false to disable ENABLE_NETBOX_MCP=true # Set to false to disable ENABLE_CATC_MCP=true # Set to false to disable ENABLE_IOS_XE_MCP=false # Disabled - won't start ENABLE_THOUSANDEYES_MCP=true ENABLE_ISE_MCP=true ENABLE_SPLUNK_MCP=false # Disabled - won't start ENABLE_PROMETHEUS_MCP=true # netops-stack metrics ENABLE_CLICKHOUSE_MCP=true # netops-stack syslog ENABLE_GITLAB_MCP=true # CI/CD orchestration ```

Best Practices: - Set ENABLE_*_MCP=false for servers you don't use - Only configure credentials for enabled servers - Use deployment profiles (below) to start specific groups - Reduces resource usage and attack surface

2. Copy and use the override configuration

cp docker-compose.override.yml.example docker-compose.override.yml

Cursor IDE Configuration

Create or update ~/.cursor/mcp.json:

{
  "mcpServers": {
    "Meraki-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8000/mcp",
      "timeout": 60000
    },
    "NetBox-MCP-Server": {
      "transport": "http", 
      "url": "http://localhost:8001/mcp",
      "timeout": 60000
    },
    "Catalyst-Center-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8002/mcp", 
      "timeout": 60000
    },
    "IOS-XE-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8003/mcp",
      "timeout": 60000
    },
    "ThousandEyes-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8004/mcp",
      "timeout": 60000
    },
    "ISE-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8005/mcp",
      "timeout": 60000
    },
    "Splunk-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8006/mcp",
      "timeout": 60000
    },
    "Prometheus-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8007/mcp",
      "timeout": 60000
    },
    "ClickHouse-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8008/mcp",
      "timeout": 60000
    },
    "GitLab-MCP-Server": {
      "transport": "http",
      "url": "http://localhost:8009/mcp",
      "timeout": 60000
    }
  }
}

LibreChat Configuration

Add to your librechat.yaml:

mcpServers:
  Meraki-MCP-Server:
    type: streamable-http
    url: http://meraki-mcp-server:8000/mcp
    timeout: 60000
  Netbox-MCP-Server:
    type: streamable-http
    url: http://netbox-mcp-server:8001/mcp
    timeout: 60000
  CatC-MCP-Server:
    type: streamable-http
    url: http://catc-mcp-server:8002/mcp
    timeout: 60000
  IOS-XE-MCP-Server:
    type: streamable-http
    url: http://ios-xe-mcp-server:8003/mcp
    timeout: 60000
  ThousandEyes-MCP-Server:
    type: streamable-http
    url: http://thousandeyes-mcp-server:8004/mcp
    timeout: 60000
  ISE-MCP-Server:
    type: streamable-http
    url: http://ise-mcp-server:8005/mcp
    timeout: 60000
  Splunk-MCP-Server:
    type: streamable-http
    url: http://splunk-mcp-server:8006/mcp
    timeout: 60000
  Prometheus-MCP-Server:
    type: streamable-http
    url: http://prometheus-mcp-server:8007/mcp
    timeout: 60000
  ClickHouse-MCP-Server:
    type: streamable-http
    url: http://clickhouse-mcp-server:8008/mcp
    timeout: 60000
  GitLab-MCP-Server:
    type: streamable-http
    url: http://gitlab-mcp-server:8009/mcp
    timeout: 60000

🌐 Server Endpoints

Verify endpoints are accessible

curl http://localhost:8000/mcp

🔗 netops-stack Integration

This MCP suite integrates with netops-stack - an observability and orchestration platform for network automation featuring:

• gNMIc - gNMI streaming telemetry collection
• Prometheus - Metrics storage and querying
• ClickHouse - Syslog and log storage via Vector
• Grafana - Visualization dashboards
• GitLab CI/CD - Ansible pipeline orchestration

Use the netops-stack profile to start MCP servers optimized for netops-stack integration:

./deploy.sh start netops-stack   # Starts MCP Servers for: ClickHouse, GitLab, IOS-XE, NetBox, Prometheus

🧩 Solution Components

External Network Integration (for LibreChat)

To integrate with LibreChat or other services on an external Docker network:

```bash

🌐 MCP Client Integration

Development Workflow

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Make your changes
4. Test with docker-compose up -d --build
5. Commit your changes (git commit -m 'Add some amazing feature')
6. Push to the branch (git push origin feature/amazing-feature)
7. Open a Pull Request

🔧 Troubleshooting