🤖 Cisco Nexus Dashboard MCP Server

Table of Contents

1. Overview
2. Features
3. Architecture
4. Requirements
5. Available MCP Tools
6. Transport
7. Installation
8. Configuration
9. Running the Server
10. Usage Example: VS Code

Overview

This MCP server acts as a lightweight middleware layer between Cisco Nexus Dashboard and any MCP-compatible client. It allows an LLM to query fabric health, anomalies, compliance, switches, interfaces, policies, VRFs/networks, and orchestration templates — all through 110 structured MCP tools.

This is a community based project and is not officially supported by any Cisco Business or Technical entity.

Features

• 110 tools across 5 ND modules — Analyze, Infrastructure, Manage, OneManage, Orchestrator
• Data-driven registration — tools sourced from a curated JSON shortlist, not hardcoded
• Streamable HTTP transport — runs as a persistent HTTP server on port 8005
• Full ND 4.2.1 coverage — anomaly analysis, compliance, fabric management, switch inventory, VRFs, networks, L4-L7 services, and orchestration
• Raw API passthrough — nd_api_request tool gives direct access to any ND endpoint
• No hardcoded credentials — all secrets sourced from .env
• Token-based authentication — automatically re-authenticates on session expiry

Architecture

OpenAPI Specs (5 files)
        │
        ▼
create_optimised_urls_4_2.py
        │
        ├─► nd_4.2.1_urls_optimised.json   (1371 endpoints, full catalogue)
        └─► nd_4.2.1_urls_shortlisted.json (107 curated tools)
                        │
                        ▼
              nd_4_2_1_server.py  (FastMCP, port 8005)
                        │
              http://localhost:8005/mcp
                        │
              MCP Client (VS Code Copilot, Claude Desktop, etc.)

Module → Base URL mapping:

Requirements

• Python 3.10+
• Access to a Cisco Nexus Dashboard 4.2.1 cluster
• ND user with read and/or write privileges

Python dependencies (requirements.txt):

Available MCP Tools

Special Tools (3 tools — always available)

Analyze Module (12 tools)

Infrastructure Module (14 tools)

Manage Module (36 tools)

Key tools include:

OneManage Module (10 tools)

Orchestrator Module (8 tools)

Transport

The server runs in streamable-http mode (FastMCP 3.2.0+) on port 8005:

http://0.0.0.0:8005/mcp

This is a persistent HTTP server — start it once and leave it running. VS Code connects to it via the mcp.json HTTP entry.

Installation

git clone <your-repo-url>
cd cd cisco_nd_mcp_commmunity

# Install dependencies
pip install -r requirements.txt

Configuration

Create a .env file in cisco_nd_mcp_community/:

NDFC_URL=https://<your-nd-ip>
USERNAME=<your-username>
PASSWORD=<your-password>
URLS_PATH=nd_4.2.1_urls_shortlisted.json
MCP_TRANSPORT=streamable-http
MCP_HOST=0.0.0.0
MCP_PORT=8005
LOG_LEVEL=INFO
HTTP_TIMEOUT=30

Security note: Never commit .env to source control — add it to .gitignore.

Running the Server

Start (recommended)

python3 /scripts/server.py

Or with the venv activated:

cd cisco_nd_mcp_community/scripts
python3 server.py

The server prints startup confirmation:

INFO  Registered 107 named tools + 3 special tools (110 total)
INFO  Nexus Dashboard 4.2.1 MCP Server
INFO    ND cluster  : https://10.x.x.x
INFO    Transport   : streamable-http
INFO    Endpoint    : http://0.0.0.0:8005/mcp

Run in background

python3 server.py &

Usage Example: VS Code

Step 1 — Add to mcp.json

Open your VS Code user mcp.json (Cmd+Shift+P → Open MCP Servers Configuration) and add:

{
  "servers": {
    "nd-4.2.1-mcp-local": {
      "type": "http",
      "url": "http://localhost:8005/mcp"
    }
  }
}

Step 2 — Start the Server

Start nd-4.2.1-mcp-local (see above), then click Restart on the nd-4.2.1-mcp-local entry in the MCP Servers panel.

Step 3 — Chat with your Nexus Dashboard

Open GitHub Copilot Chat (Cmd+Shift+I) and try natural language prompts:

📋 Fabric overview:

List all fabrics managed by Nexus Dashboard

Show me the ND cluster version and health status

What smart licensing is active?

🚨 Anomalies & compliance:

Show me all critical anomalies across fabrics

Which fabrics have the highest anomaly score?

List compliance anomaly rules for the fabric

🏗️ Manage & deploy:

List all tenants in the manage module

Show me pending configuration changes for the fabric

Deploy networks to the fabric

🔍 Inventory & interfaces:

List all switches bootstrapped in the fabric

Show me all logical links

Get interface groups for the fabric

The LLM will automatically select the right MCP tool(s), call the APIC REST API, and return structured results in the chat window.

LICENSE

This sample code is licensed under the Apache 2.0 license as per the LICENSE file attached