Opendatasoft MCP Server

Opendatasoft MCP Server

A Model Context Protocol (MCP) server that provides tools for interacting with the Opendatasoft Explore API v2.1, enabling AI assistants like Claude to search, query, and analyze open datasets.

Features

• Dataset Discovery: Search and browse datasets by keywords, publishers, and themes
• Dataset Exploration: View schemas, metadata, and sample records
• Data Querying: Execute ODSQL queries with filtering, sorting, and aggregation
• Data Analysis: Generate statistics, analyze fields, and visualize distributions
• Data Export: Generate export URLs for various formats (CSV, JSON, GeoJSON, etc.)

Installation

Requirements

• Python 3.10 or later
• Model Context Protocol (MCP) SDK 1.2.0 or later
• Claude for Desktop or another MCP-compatible client

Installing from Source

1. Clone the repository:

   git clone https://github.com/your-username/opendatasoft-mcp-server.git
   cd opendatasoft-mcp-server
   

2. Create a virtual environment and install dependencies:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -e .
   

Configuration

The server can be configured using environment variables:

• ODS_BASE_URL: Base URL for the Opendatasoft domain (default: "https://documentation-resources.opendatasoft.com")
• ODS_API_KEY: API key for authenticated requests (optional)

Usage with Claude for Desktop

1. Make sure you have Claude for Desktop installed. You can download it from claude.ai/download.

2. Configure Claude for Desktop to use this MCP server by adding it to your Claude for Desktop configuration file:

   • On macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • On Windows: %APPDATA%\Claude\claude_desktop_config.json

   {
     "mcpServers": {
       "opendatasoft": {
         "command": "/path/to/venv/bin/python",
         "args": ["-m", "src.main"],
         "env": {
           "ODS_BASE_URL": "https://documentation-resources.opendatasoft.com",
           "ODS_API_KEY": "your-api-key-if-needed"
         }
       }
     }
   }
   

3. Restart Claude for Desktop.

4. You can now use the Opendatasoft MCP server in your conversations with Claude.

Available Tools

Catalog Tools

• search_datasets: Search for datasets by keyword
• get_dataset_info: Get detailed information about a specific dataset
• list_datasets_by_publisher: List datasets from a specific publisher
• list_dataset_fields: List all fields in a dataset with their types and descriptions

Query Tools

• get_dataset_records: Get records from a dataset with optional filtering and sorting
• get_dataset_aggregates: Get aggregated data from a dataset using ODSQL aggregation functions
• facet_analysis: Analyze facet values distribution for a dataset
• search_dataset_records: Search for specific records within a dataset
• get_export_url: Get a URL for exporting dataset records in various formats

Analysis Tools

• summarize_dataset: Generate a comprehensive summary of a dataset
• analyze_numeric_field: Analyze a numeric field, including min, max, average, and distribution
• analyze_text_field: Analyze a text field, including value frequency
• analyze_date_field: Analyze a date field, including range, distribution by year/month
• generate_dataset_statistics: Generate comprehensive statistics for all fields in a dataset

Example Queries for Claude

Here are some example queries you can ask Claude while using this MCP server:

• "Find datasets related to transportation."
• "Show me datasets published by the World Food Programme."
• "What are the fields in the 'world-administrative-boundaries' dataset?"
• "Get 5 records from the 'gold-prices' dataset."
• "Count the number of cities per country in the 'geonames-all-cities-with-a-population-1000' dataset."
• "Analyze the 'population' field in the 'world-administrative-boundaries' dataset."
• "What's the distribution of records by year in the 'gold-prices' dataset?"
• "Generate a CSV export URL for the 'gold-prices' dataset with prices sorted by date."

Understanding ODSQL

Many of the tools in this MCP server use the Opendatasoft Query Language (ODSQL) for filtering, aggregating, and sorting data. Here are some basic examples:

Select Clause

Choosing which fields to return:

select=field1, field2, field3

Aggregation:

select=count(*) as total, avg(field) as average

Where Clause

Filtering records:

where=field > 100
where=date_field >= date'2020-01-01'
where=text_field like "Paris"

Full-text search:

where=search(field, "keyword")

Group By Clause

Grouping results:

group_by=field
group_by=year(date_field)

Order By Clause

Sorting results:

order_by=field ASC
order_by=field DESC

For more details on ODSQL syntax, see the Opendatasoft documentation.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Opendatasoft for providing the Explore API
• Model Context Protocol (MCP) for enabling AI assistants to interact with tools
• Claude for the AI assistant capabilities