AGESIC and OKFN's Traceable AI for Uruguay's Open Data Catalogue

Uruguay's open-data catalogue (catalogodatos.gub.uy) holds thousands of government datasets, but citizens without technical skills struggle to find and interpret them, and generic AI chatbots risk answering from ungrounded or fabricated information rather than the real data.
AGESIC — the Agency for Electronic Government and the Information and Knowledge Society, under Uruguay's Office of the President — partnered with the Open Knowledge Foundation's (OKFN) Tech Team to prototype a Model Context Protocol (MCP) server. It lets users ask natural-language questions and receive answers traced directly to the underlying open datasets, instead of free-form generated text. The tool is designed to work with any CKAN-based data portal, is funded by the Patrick J. McGovern Foundation, and follows the open, reusable model promoted by the Digital Public Goods Alliance.
Announced in June 2026 as part of Uruguay's National Artificial Intelligence Strategy 2024-2030 and National Data Strategy 2030, the project is explicitly an early pilot: OKFN's own announcement reports no usage metrics or performance figures, and a parallel prototype is being tested in Brazil with the Office of the Comptroller General.
Because the tool is so new, its real-world accuracy, adoption and error rate remain unmeasured. It is included here as an early, transparently-designed model for verifiable AI access to public data — not yet as a proven success.

Read the full analysis: https://fairlac.iadb.org/en/hub/uruguay