FIWARE MCP Server

What is FIWARE MCP Server?

The FIWARE MCP Server is a first implementation of a Model Context Protocol (MCP) server that provides a bridge between a FIWARE Context Broker and other services. It implements basic operations for interacting with a FIWARE Context Broker, such as version checking, querying, and publishing/updating entities. This server is intended for developers who need to integrate FIWARE Context Broker capabilities into MCP-compatible clients (e.g., Claude Desktop).

How to use FIWARE MCP Server?

Install the dependencies (pip install -r requirements.txt), then run the server in STDIO mode (default) with python server.py or in HTTP mode with --http flag. For Claude Desktop integration, configure the MCP client to launch the server via STDIO using either uv or the Python interpreter directly (see README for full JSON configuration). The server exposes three tools: CB_version (check broker version), query_CB (query entities), and publish_to_CB (publish/update entities). In HTTP mode, use --host and --port options; the MCP endpoint is available at /mcp. To expose the server to external LLMs, use ngrok to create a public HTTPS tunnel.

Key features of FIWARE MCP Server

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

Use cases of FIWARE MCP Server

• Check the version of a running FIWARE Context Broker
• Query entities (e.g., temperature sensors) from the Context Broker
• Publish or update NGSI-LD entity data in the Context Broker
• Integrate Context Broker interactions into MCP-based AI assistants (e.g., Claude Desktop)
• Expose Context Broker operations to external LLMs via HTTP mode and ngrok

FAQ from FIWARE MCP Server

What are the runtime requirements?

Python 3.7 or higher, pip, and access to a FIWARE Context Broker instance. Required Python packages include requests>=2.31.0 and fastmcp>=0.3.0 (listed in requirements.txt).

How do I run the server in HTTP mode?

Use the --http flag: python server.py --http. Optionally specify host and port: python server.py --http --host 127.0.0.1 --port 5001.

How can I expose the server to external APIs (e.g., OpenAI Responses API)?

Run the server in HTTP mode, then use ngrok to create a public HTTPS tunnel pointing to the server’s host/port. The MCP endpoint will be reachable at {PUBLIC_URL}/mcp.

What tools are available and what parameters do they accept?

• CB_version: parameters address (default "localhost") and port (default 1026). Returns version info.
• query_CB: parameters address, port, and query (default ""). Returns query results.
• publish_to_CB: parameters address, port, and entity_data (required dictionary). Returns operation status.

How is error handling done?

The server includes comprehensive error handling for network connectivity issues, invalid Context Broker responses, malformed entity data, and server shutdown.