MCP server for IP Fabric

MCP server to interact with IP Fabric via the python SDK, initially inspired by the MCP Server for Obsidian.

⚠️ Disclaimer — Unofficial & Experimental

This is NOT an official IP Fabric product or project. This MCP server was built as a personal experiment to explore, test, and learn about the Model Context Protocol (MCP) and how it can interact with IP Fabric through its Python SDK. It is not developed, maintained, endorsed, or supported by IP Fabric in any official capacity. 🚀 An official IP Fabric MCP server is currently being developed and tested by the IP Fabric team. If you are looking for an official, production-ready integration, please reach out to your Solution Architect for the latest updates on availability and features. ### What this means for you: - Do not use this in production environments. This project may contain bugs, incomplete features, or breaking changes at any time. - No guarantees. There is no warranty, SLA, or official support associated with this project. - No affiliation. This repository is not affiliated with, endorsed by, or connected to IP Fabric's official MCP server efforts. - Use at your own risk. You are responsible for any consequences of using this code in your environment. If you have questions about IP Fabric's official MCP server, please contact your Solution Architect or reach out to the IP Fabric team directly. _This project exists purely for educational and experimental purposes. Contributions and feedback are welcome, but please set your expectations accordingly!_ 🧪

Components

Tools

The server implements multiple tools to interact with IP Fabric:

ipf_get_filter_help: Provides help information for using filters in queries
ipf_get_snapshots: Lists all available snapshots in IP Fabric
ipf_set_snapshot: Sets the active snapshot for subsequent queries
ipf_get_devices: Gets device inventory data with optional filters
ipf_get_interfaces: Gets interface inventory data with optional filters
ipf_get_hosts: Gets host inventory data with optional filters
ipf_get_sites: Gets site inventory data with optional filters
ipf_get_vendors: Gets vendor inventory data with optional filters
ipf_get_routing_table: Gets routing table data with optional filters
ipf_get_managed_ipv4: Gets managed IPv4 data with optional filters
ipf_get_vlans: Gets VLAN data with optional filters
ipf_get_neighbors: Gets neighbor discovery data with optional filters
ipf_get_available_columns: Gets available columns for specific table types
ipf_get_connection_info: Gets IP Fabric connection information and status

Example prompts

It's good to first instruct Claude to use IP Fabric. Then it will always call the tools.

Use prompts like this:

"Show me all available snapshots in IP Fabric"
"Set the snapshot to the latest one and show me all devices"
"Get all Cisco devices from the inventory"
"Show me all interfaces on router 'core-01'"
"Find all routes to 192.168.1.0/24"
"Get devices with hostname containing 'switch'"
"Show me the routing table for devices in site 'headquarters'"
"What columns are available for the devices table?"

Configuration

Environment Variables

The server uses environment variables for configuration. Copy the .env.sample file to .env and update the values accordingly:

cp .env.sample .env

Required Environment Variables

IP Fabric Configuration

# IP Fabric Configuration
IPF_URL=https://ipfabric-server.domain
IPF_TOKEN=your_api_token_here

AI Model Configuration

Choose one of the following AI providers and set the AI_MODEL and AI_API_KEY variables accordingly.

Here are some examples:

# OpenAI (default)
AI_MODEL="gpt-4o"
AI_API_KEY=sk-proj-xxx

# Anthropic
AI_MODEL="anthropic/claude-sonnet-4-0"
AI_API_KEY=sk-ant-api...

# Google Gemini
AI_MODEL="gemini/gemini-2.5-flash"
AI_API_KEY=xxx

Optional Environment Variables

LangSmith Tracing (Optional)

# Enable tracing with LangSmith
LANGSMITH_TRACING=true
LANGSMITH_ENDPOINT="https://eu.api.smith.langchain.com"
LANGCHAIN_API_KEY=lsv2_xx_123..._123...
LANGSMITH_PROJECT="ipf-mcp-2025-07"

Configuration Methods

Add to server config (preferred)

    {
      "mcp-ipf": {
        "command": "/path/to/uv",
        "args": [
          "--directory",
          "<path_to_this_repo>",
          "run",
          "--env-file",
          ".env",
          "src/mcp_ipf/server.py"
        ],
        "env": {
          "IPF_TOKEN": "<your_api_token_here>",
          "IPF_URL": "<your_ip_fabric_host>",
          "AI_MODEL": "<your_ai_model>",
          "AI_API_KEY": "<your_ai_api_key>"
        }
      }
    }

Sometimes Claude has issues detecting the location of uv / uvx. You can use which uvx to find and paste the full path in above config in such cases.

Use .env file in the working directory with the required variables (copy from .env.sample):

    cp .env.sample .env
    # Edit .env with your actual values

Quickstart

Prerequisites

IP Fabric API Access

You need IP Fabric API access with a valid API token. Get this from your IP Fabric instance:

Log into your IP Fabric instance
Go to Settings → API tokens
Create a new API token
Copy the token for use in configuration

Claude Desktop

On MacOS: ~/Library/Application\ Support/Claude/claude_desktop_config.json

!!! note it's recommended to use the full path to uv in the configuration, as sometimes Claude has issues detecting the location of uv. Use which uv to find the full path and paste it in the command field of the configuration.

<details> <summary>Development/Unpublished Servers Configuration</summary>

    {
      "mcpServers": {
        "mcp-ipf": {
          "command": "/path/to/uv",
          "args": [
            "--directory",
            "<path_to_this_repo>",
            "run",
            "--env-file",
            ".env",
            "src/mcp_ipf/server.py"
          ],
          "env": {
            "IPF_TOKEN": "<your_api_token_here>",
            "IPF_URL": "<your_ip_fabric_host>"
          }
        }
      }
    }

</details>

Raycast AI - MCP Servers

Open Raycast, type mcp and select Install Server

!Screenshot of Raycast search after typing mcp

Fill the form with the following details:

command: /path/to/uv

If unsure, use which uv to find the full path.

arguments: --directory <path_to_this_repo> run --env-file .env src/mcp_ipf/server.py

!Screenshot of Raycast MCP server installation process

Now you can install the server with ⌘ + ⏎

Using the CLI

To use the CLI application, after setting up your environment variables:

uv run python cli_app.py

Using Streamlit

...coming soon...

Development

Project Structure

playground-mcp-ipf/
├── src/
│   └── mcp_ipf/
│       ├── __init__.py      # Package entry point
│       ├── server.py        # MCP server implementation
│       ├── tools.py         # Tool handlers
├── .env                      # Environment variables (copy from .env.sample)
├── .env.sample              # Sample environment variables
├── cli_app.py               # CLI application using the MCP server
├── pyproject.toml
└── README.md

Running

Run the server directly during development:

uv run mcp-ipf

Adding New Tools

To add new IP Fabric tools:

Create a new tool handler class in tools.py
Add the tool class to the tool_classes list in server.py
The tool will be automatically registered and available

Supported AI Models

The server supports multiple AI providers through LiteLLM:

OpenAI: gpt-4o, gpt-4o-mini, gpt-3.5-turbo, etc.
Anthropic: anthropic/claude-sonnet-4-0, anthropic/claude-haiku-3-5, etc.
Google Gemini: gemini/gemini-2.5-flash, gemini/gemini-pro, etc.

See the respective provider documentation for full model lists:

Troubleshooting

Common Issues

Connection errors: Verify your IPF_URL and IPF_TOKEN are correct
SSL certificate issues: Check your IP Fabric server's SSL configuration
Permission errors: Ensure your API token has sufficient permissions in IP Fabric
Snapshot issues: Use ipf_get_snapshots to see available snapshots, then ipf_set_snapshot to select one
Environment variable issues: Ensure your .env file is properly configured and accessible

Getting Help

Check the server logs: tail -f ~/Library/Logs/Claude/mcp-server-mcp-ipf.log
Use the MCP Inspector for debugging
Verify your IP Fabric API token has the necessary permissions
Ensure your IP Fabric instance is accessible from your machine

IP Fabric MCP Server