Loomal

Best Open Data MCP servers for AI agents.

Government statistics, parliamentary records, and national data portals — exposed as MCP tools so agents can query official sources directly instead of scraping them.

Open data is the rare category where the underlying content is genuinely free: governments and public institutions publish statistics, budgets, and records under open licenses. The problem has never been access — it's that every portal speaks a different API dialect. Open Data MCP servers solve that by wrapping each portal in a standard tool interface an AI agent can call without custom integration code.

The servers indexed here range from portal-software adapters like ckan-mcp-server, which works against any CKAN-based portal, to single-source wrappers for specific national institutions — Statistics Sweden, FEMA, the Norwegian and Dutch parliaments.

What Open Data MCP servers do

These servers translate public-sector data APIs into MCP tools. ckan-mcp-server is the most general of the group: CKAN powers thousands of open data portals worldwide (including many city and national portals), so one server gives an agent dataset search and retrieval across any of them. Others go deep on a single source — scb-opendata-mcp exposes Statistics Sweden's tables, openfema covers US Federal Emergency Management Agency disaster data, and foehn serves Swiss meteorological data from MeteoSwiss.

A distinct sub-cluster covers parliamentary records: Storting No, Tweedekamer Nl, and Riksdagen Se wrap the open data APIs of the Norwegian, Dutch, and Swedish parliaments respectively. For an agent doing legislative research, these turn 'find every vote on topic X' from a scraping project into a single tool call.

What to look for when choosing one

Coverage shape matters most. A portal-software adapter like ckan-mcp-server is broad but generic — you get dataset search and metadata, and the agent still has to interpret whatever CSV or JSON the portal returns. A source-specific server like scb-opendata-mcp or openfema can offer typed, purpose-built tools because the author knows the data model.

Also check locale and language: many of these servers (Italy OpenData, mcp-new-caledonia, Nsi Bg) serve data in the source country's language and administrative geography. That's a feature if your agent works in that market and a trap if it doesn't. Finally, read the upstream license — open data is usually attribution-licensed, and your agent's output may need to carry that attribution.

How agents actually use them

The dominant pattern is grounding: an agent answering a question about unemployment, disaster declarations, or parliamentary votes calls the relevant server, gets official numbers, and cites them — instead of relying on whatever was in its training data. Because the sources are authoritative, these tools are popular in research and journalism workflows where 'where did this number come from' must have a real answer.

Cross-source comparison is the second pattern. An agent with the Swedish, Norwegian, and Dutch parliament servers attached can compare legislative activity across countries in one conversation — something that previously meant three separate API integrations.

Free data, real serving costs

The data is free; serving it reliably isn't. Most servers in this category are open source and free to self-host, but a hosted endpoint that stays up, caches portal responses, and handles upstream rate limits costs the maintainer real money. That's where x402 pay-per-call pricing fits: a maintainer can claim their listing on Loomal, set a per-call price (minimum $0.01 in USDC, settled on Base), and let agents pay for the hosted convenience while the code stays open.

Each listing below links to its marketplace page, where claimed servers show live pricing alongside the tool list.

Frequently asked questions

What are the best Open Data MCP servers?

It depends on the source you need. ckan-mcp-server is the broadest pick because it works against any CKAN-based portal; scb-opendata-mcp, openfema, and foehn are stronger choices when you need one specific institution's data with purpose-built tools. The Loomal index currently tracks 21 live servers in this category.

Are Open Data MCP servers free to use?

The underlying data is openly licensed and the servers themselves are generally open source, so self-hosting is free apart from your own infrastructure. Hosted endpoints are a different matter — some maintainers charge per call via x402 to cover serving costs, which agents pay automatically in USDC.

Why use an MCP server instead of calling the portal API directly?

Every open data portal has its own API conventions, pagination, and query language. An MCP server normalizes that into tools an agent can discover and call without custom code, and the maintainer keeps the integration working when the portal changes. For one-off scripts a direct API call is fine; for agents, the standard interface wins.

How do I get my own Open Data MCP server listed here?

Publish it to the official MCP registry — Loomal indexes from there. Once it appears, claim the listing by verifying ownership of the GitHub repository, after which you can edit the listing and optionally set per-call pricing in the Loomal console.

Run a Open Data MCP server?

Claim your listing, set a per-call USDC price, and let AI agents pay for every call over x402.

List it on Loomal