A few months ago, Luis Figueroa and I started building EcuDataMCP, mostly out of frustration with how difficult it can be to work with Ecuadorian public data.
Depending on what you need, you might end up searching a CKAN catalogue, downloading an Excel file from the SRI, figuring out an undocumented endpoint, or navigating a government website that was not really designed for programmatic access. Once you find the data, there is usually another round of cleaning before you can actually use it.
EcuDataMCP is a common interface for that mess, and it’s proven to be very successful.
What EcuDataMCP actually does
EcuDataMCP is an open-source Model Context Protocol (MCP) server. In practical terms, it gives tools such as Claude Code, Codex, Cursor, and other MCP-compatible clients a way to query Ecuadorian public data directly.
Instead of asking an AI assistant to search the web for a dataset and hope it finds the right file, you can give it a specific tool for the job. Those kinds of searches are often not very successful for Ecuadorian data.
Some of the main sources currently exposed through the MCP are:
| Source | What it exposes |
|---|---|
Open Data (datosabiertos.gob.ec and other CKAN portals) |
National and municipal data catalogues, resource previews, metadata, and tabular DataStore queries without downloading the underlying file |
| SRI | Tax collection, RUC registrations, sales and purchases, vehicles, and other statistical datasets published outside the main CKAN portal |
| Gob.ec | Government procedures, requirements, costs, institutions, and regulations linked to the Official Registry |
| SERCOP (OCDS) | Public procurement procedures, buyers, tenders, awards, and contracts |
| SGR | Emergency events, SITREP reports, risk-management documents, and tsunami early-warning stations |
| IG-EPN | Recent earthquakes and the archive of seismic and volcanic reports |
| ANDA / INEC | Survey and census catalogue, metadata, and links to downloadable microdata |
| INEC | Statistical publications, Ecuador en Cifras, Census 2022 resources, and other public data products |
| BCE | Macroeconomic series, IEM tables, publications, remittances, national accounts, external-trade prices, and daily indicators |
| Superintendencia de Compañías | Directory of more than 226,000 companies, company records, financial statements and ratios, rankings, and external auditors |
There are 113 tools at this point. Some search datasets, some query individual resources, and others talk directly to a particular source. The server now reaches well beyond the sources above, including data from agriculture, labour, education, aviation, telecommunications, banking, energy, social protection, weather, and other areas.
That means I can ask for recent earthquakes above magnitude 4, search what tax data the SRI publishes, retrieve a BCE series, or look up firms without first figuring out where each institution happens to store that information.
Building around messy public data
There is no single Ecuadorian government data API behind EcuDataMCP. There are multiple APIs, and a lot more beyond them. Some sources have reasonably clean APIs (the Central Bank, for example). Others expose CKAN resources, that is, well-structured government data portals. A few are essentially endpoints sitting behind websites that were primarily built for people clicking around in a browser (INEC, I’m looking at you).
EcuDataMCP wraps those different systems into a set of tools with predictable inputs and outputs. Once that layer exists, the client using it does not need to know nearly as much about how each source works.
There are still limitations. Some Ecuadorian government sources restrict or behave inconsistently with traffic from outside the region, so a Latin American VPN may be necessary for some tools when running the MCP from abroad.
Sismicidad Ecuador
I built Sismicidad Ecuador because I wanted something more useful to demonstrate what this setup makes possible. The dashboard uses EcuDataMCP’s search_sismos tool, which retrieves recent seismic events from the Instituto Geofísico de la Escuela Politécnica Nacional. From there, I built a single-page interface for exploring the events geographically and over time.
It includes an interactive map using Ecuador’s actual provincial and cantonal boundaries, a timeline of recent events, filters for magnitude and depth, and an event panel that stays synchronized with the map.
I built the dashboard with Claude Code, starting from a rough version and iterating on it repeatedly. The first versions looked exactly like what they were: a dataset with some controls attached to it. I ended up spending much more time on the map, information hierarchy, filters, typography, and interaction between components than on the initial data retrieval.
This is a good example of what the MCP changes. Once the data-access problem becomes trivial, a larger share of the work can go into what you actually want to do with the data.
Try it
You can open Sismicidad Ecuador directly in your browser and explore the latest events, or try it right here:
The full EcuDataMCP code is also available on GitHub if you want to run it, inspect the tools, or build something else on top of them. The easiest way to use it is to point your AI agent to it (Codex, Claude Code, Cursor, etc.).
Sismicidad Ecuador is also available on my dashboards page alongside my other data-visualization projects.