Our data shows its work — and now machines can read it

Quantamentry — per-indicator provenance on every country page, a public MCP server, and machine-readable surfaces.

"Trust the data" is easy to say and hard to check. This release is about making it checkable — by people and by the tools they increasingly send to read the web on their behalf. Two things shipped: every dimension on a country page now shows exactly what it's built on, and the scores are now queryable by AI agents through a public MCP server. Both come from the same conviction that runs through everything here — if a number can't show its source, you shouldn't have to trust it.

Every dimension now shows its source

Open any country page and each of the seven dimensions is expandable. Click one and it reveals its provenance: the source that produced the reading, the "as of" date of the underlying data, and the coverage tier behind it. If a dimension has no current data, it says so; if it doesn't apply to that country, it says that too. No client-side JavaScript is involved — it's a plain expandable block, so it works everywhere and loads instantly.

The same discipline runs through the rest of the page. The key-indicators table labels each row with the exact source that produced that value — Eurostat, Ember, IMF PortWatch, World Bank via NASA nightlights, whichever series actually backs the number — and its date, right there in the row. There is no "data provided by various sources" hand-wave. If you can't trace a number back to a source, we don't think you should trust it, and now you don't have to take our word for where it came from: it's printed next to the value.

This is the retail version of the licensing discipline the platform has always run on. We ingest plenty of series that can't be redistributed raw — so what leaves the building is derived indicators, and the provenance labels tell you which source each derivation sits on top of.

A public MCP server for agents

The bigger structural change is that the scores are now readable by AI agents directly. There is a public, read-only MCP server — the Model Context Protocol, the emerging standard for letting language-model agents call tools — live at quantamentry.com/api/mcp. Point an MCP-capable agent at that endpoint and it can query our data without scraping a single HTML page.

It exposes six read-only tools:

• get_country_score — the current composite and all seven dimensions for one country, with per-dimension provenance and recent deltas.
• list_rankings — countries ranked by composite score, optionally filtered by region.
• get_score_history — the 90-day daily composite history for a country.
• get_coverage — per-source freshness and the per-tier country counts.
• get_methodology — the current methodology version and a short summary of what the score is, and isn't.
• compare_countries — two to six countries side by side across all seven dimensions.

A deliberate note on what it is not: it is read-only and unauthenticated, and every tool reads only data that is already public on the site. There is nothing behind it that isn't already on a page a person can open. It speaks plain JSON-RPC 2.0 over HTTP POST, with a courtesy per-IP rate limit — no accounts, no keys, no session dance to get a number out of it.

llms.txt and machine-readable pricing

Rounding out the agent-readable surface: there's a /llms.txt that summarises the whole site for language models that ground on it, a /pricing.md that mirrors the pricing page as plain Markdown, and a /api/product-catalog that serves the same tiers as JSON. The machine-readable pricing surfaces are generated from one source of truth, so they can't drift from what a human reads on the page — the JSON an agent sees carries the same numbers a person does.

None of this is a metaphor or a roadmap item. The endpoints are live now, and they exist because the audience that cares most about a fully-cited public-data stack — data and OSINT people, indie quants, and the agents they're starting to build — should be able to consume it the way they actually work: as structured data, not as a page to parse.

See it live

Open any country page — Turkey is a good one — and expand a dimension to see its source, date, and tier. Agent builders can hit quantamentry.com/api/mcp directly, or start from /llms.txt.

If you'd rather just watch a few countries the human way, Create a free account — watch 3 countries. Free, and it takes about a minute.

Everything above surfaces our 0–100 composite scores, built entirely on free public data (World Bank, IMF, BIS, FRED, OECD, ILO, GDELT, European Central Bank), with each indicator's own source labelled on the page.

— Quantamentry, July 28, 2026