One matrix
Each cell takes the best status across all products, and links to its sources. We read two kinds of source: research and regulation, and what companies actually ship.
See what AI agents can actually do, country by country.
Status matrix
Ranking
Score 0–100 · tierBy category
0–100Newest evidence
1–6jump to a panel JKnext or previous country Escclose details
The dataset · as of
Counted cell by cell.
Compare countries
How the countries compare. One table first, then every rank.
Six charts, from the simplest to the richest.
Companies
Who makes it possible. Based where, live where.
Open source
What agents are built on. Really used, or only starred.
Free projects anyone can build an agent on, without asking a company.
Stars measure attention. Installs measure use. So we sort repos by npm, PyPI and Packagist installs a month and Docker Hub pulls, not by stars. How we tier repos
Where it comes from
Built on dated sources. When they appeared, how we score.
Method
Every status cites a source. Thin evidence looks thin.
Doubt is drawn, not hidden
When the sources are unclear, a cell takes the lower status.
Every category weighs the same
A category scores the mean of its cells' points. A country scores the mean of its categories. Live, Limited and Announced cells must cite a source, or the build fails.