Pilots with the Uruguay and Brazil governments showed how a Model Context Protocol (MCP) provides a structured way for AI to tap into up-to-date and actual data from open portals, instead of relying only on the LLM's general training. The Open Knowledge Foundation (OKC) collaborated with Brazil's Office of the Comptroller General and Uruguay’s Agency for Electronic Government and the Information and Knowledge Society to prototype an MCP bridge between LLMs and open data portals. The goal is to let citizens ask natural language questions and get answers traceable back to official datasets. The pilots shared the same architecture but focused on different datasets: Brazil’s pilot focuses on a dataset about parliamentary amendments, while Uruguay’s pilot works with the National Energy Balance, covering areas such as energy imports, generation, and installed capacity. The setup included an MCP server connecting the AI to datasets, a chat gateway, and tools explaining what questions can be asked. Both countries selected datasets based on real user demand, with partners providing frequently asked questions for testing. The MCP layer connected to the data, exposed tools, and gave the language model a structured way to get fresh data. However, the quality of answers depended heavily on the quality of descriptions around the data, as someone had to explain what the data meant, what fields represented, what units were used, and what time periods were covered.