UN statistics now sit behind an MCP connection an agent can query itself

Global statistics have been available for years and awkward to use for just as long, because each agency published in its own format with its own geography and its own timeline. On September 17, 2026, the UN system opened the UN System Data Commons, built on Google's Data Commons, which merges those datasets into one knowledge graph. For anyone writing software, the part that matters is the interface: the platform speaks the Model Context Protocol, so an agent can ask it for a number directly.

What was actually opened

The platform is open source and lives at data.un.org. It accepts questions in plain language — how access to clean water in rural areas affects school attendance, how many people gained access to electricity in the past decade — and answers with data and interactive charts. An Explore tab filters by location or by theme such as health or education, and a blog section publishes worked analyses, one of them using UNICEF data on child poverty.

The merging is the expensive part and the platform does it once for everyone. Metrics, timelines and geographic boundaries from different agencies are aligned into a single environment, which is the work analysts previously repeated by hand before any analysis began. Google states that every dataset was validated with UN system statisticians and technical experts, and that the project was supported by a Google.org grant to the UN Foundation.

How an agent plugs into it

Data Commons exposes the Model Context Protocol, which is the same interface a coding agent already uses to reach a file system or a ticket tracker. An agent pointed at this endpoint can fetch authoritative figures, join them across domains, and assemble charts, infographics or a draft report without a person exporting anything.

That changes the shape of a whole class of small tools. A dashboard that needed a scraper, a cleaning step and a scheduled job to stay current can instead ask a question at request time and get an official figure back, with the provenance attached to it.

• The data is official statistics from UN entities, not a third party's copy of them.
• The protocol is the one your agent already speaks, so no bespoke client is needed.
• Geography and time are normalized across agencies, which is where hand-built pipelines usually break.
• The launch post carries a caution about citing figures without opening the source underneath them.

What the announcement leaves open

Coverage is a stated plan rather than a finished state. The UN system says it will keep adding datasets from more entities over the coming year, aiming to include 80% of the system's statistical datasets by 2027. What is present on launch day is therefore a fraction of that, and the post does not say how large a fraction.

Rate limits and terms for automated access are not described in the announcement either. Before a production service depends on this, both are worth confirming against the platform's own documentation rather than the launch post.