FIWARE MCP Server

This is a first implementation of a FIWARE Model Context Protocol (MCP) Server that provides a bridge between the Context Broker and other services. The server implements basic operations for interacting with a FIWARE Context Broker.

Objectives

• Create a basic MCP server implementation for FIWARE
• Provide simple tools for Context Broker interaction
• Demonstrate basic intent CRUD operations with the Context Broker
• Serve as a foundation for more complex MCP implementations

Features

• Context Broker version checking
• Query capabilities for the Context Broker
• Entity publishing and updating

Prerequisites

• Python 3.7 or higher
• pip (Python package installer)
• Access to a FIWARE Context Broker instance

Installation

1. Clone this repository:

git clone 
cd FIWARE_MCP_01

2. Install the required dependencies:

pip install -r requirements.txt

Claude Desktop integration

Claude Desktop should launch the server via STDIO. Do not use “mcp install” or “mcp run” with this server.

Example configuration (adjust paths for your environment):

{
  "mcpServers": {
    "CB-assistant": {
      "command": "uv",
      "args": [
        "run",
        "--with",
        "requests>=2.31.0",
        "--with",
        "fastmcp>=0.3.0",
        "python",
        "PATH\\TO\\FIWARE-MCP-Server\\server.py"
      ]
    }
  }
}

Alternative if dependencies are installed with pip:

{
  "mcpServers": {
    "CB-assistant": {
      "command": "python",
      "args": [
        "PATH\\TO\\FIWARE-MCP-Server\\server.py"
      ]
    }
  }
}

Usage

By default, running the script starts the server in STDIO mode. Use `--http` to start an HTTP server.

• STDIO (default):

python server.py

• HTTP mode:

python server.py --http

• HTTP host/port (optional):

python server.py --http --host 127.0.0.1 --port 5001

When running with `--http`, the server binds to the provided host/port and uses stateless HTTP sessions for compatibility with streamable HTTP clients.

Redirecting with ngrok (HTTP mode)

To use LLMs via external APIs (for example the OpenAI Responses API) you may need to expose your local MCP server to the Internet. Use ngrok to create a public HTTPS tunnel to the server when running in HTTP mode.

1. Sign up at https://ngrok.com and install the ngrok client for your OS.

2. Add your ngrok auth token to the client (follow ngrok's OS-specific setup). For example:

3. Start your MCP server in HTTP mode (if not already running). Example:

4. Start an ngrok tunnel that forwards to your HTTP server:

(replace host/port if you configured them differently)

5. Once ngrok is running it will display one or more public forwarding URLs. The MCP endpoint will be reachable at:

Note: Keep your auth token secure. The public URL stays active while ngrok is running and will cease to be reachable when you stop the tunnel.

Available Tools

1. CB_version

2. query_CB

3. publish_to_CB

Example Usage

# Example entity data
entity_data = {
    "id": "urn:ngsi-ld:TemperatureSensor:001",
    "type": "TemperatureSensor",
    "temperature": {
        "type": "Property",
        "value": 25.5
    },
    "@context": "https://uri.etsi.org/ngsi-ld/v1/ngsi-ld-core-context.jsonld"
}

# Publish to Context Broker
result = publish_to_CB(entity_data=entity_data)

Configuration

The server can be configured by modifying the following parameters in `server.py`:

• Host address (HTTP mode only)
• Port number (HTTP mode only)
• Timeout settings

Error Handling

The server includes comprehensive error handling for:

• Network connectivity issues
• Invalid responses from the Context Broker
• Malformed entity data
• Server shutdown

Contributing

Feel free to submit issues and enhancement requests!

License

This project is licensed under the Apache License 2.0.