AI assistant access to CKAN open data portals
This server implements the Model Context Protocol (MCP), allowing AI assistants like Claude, ChatGPT, and GitHub Copilot to search datasets, query DataStore records, and explore portal metadata on any CKAN instance.
Add the server from your terminal:
Then ask Claude anything about the portal's data.
Add the following to your claude_desktop_config.json:
Config file location:
~/Library/Application Support/Claude/claude_desktop_config.json (macOS),
%APPDATA%\Claude\claude_desktop_config.json (Windows).
Restart Claude Desktop after editing.
Point any MCP-compatible client at the Streamable HTTP endpoint:
| Tool | Description |
|---|---|
dataset_search | Search datasets by keyword, filter, or facet |
dataset_show | Get full metadata for a specific dataset |
dataset_list | List all dataset names/IDs |
datastore_search | Query records in a DataStore resource |
datastore_fields | Get field definitions for a DataStore resource |
resource_show | Get metadata for a specific resource |
organization_list | List all organisations |
organization_show | Get organisation details |
group_list | List all groups/themes |
group_show | Get group details |
tag_list | List all tags |
ckan-fynd connects to a CKAN portal's Action API over HTTP and translates MCP tool calls into CKAN API requests. No database, no state — just a protocol bridge between AI assistants and open data.
For portals that run CKAN 2.10+, consider ckanext-fynd instead — a CKAN extension that serves MCP from inside CKAN with native auth integration.