Back to blog

Giving Your AI Agent Live Economics Data via MCP

How to connect an AI agent to real GDP, inflation, and trade data over MCP instead of scraping pages or hand-rolling API calls.

An AI agent answering "what's Germany's current inflation rate" has three options: guess from training data that's already stale, scrape a web page and hope the table parses cleanly, or call a tool built for exactly this. MCP — the Model Context Protocol — is what makes the third option as easy as the first two, and it's why we built a hosted MCP server alongside the REST API rather than treating agents as an afterthought.

What MCP actually gives an agent

MCP standardizes how a model discovers and invokes external tools. Instead of an agent trying to parse the Germany hub page's HTML — fragile, and liable to break the next time we change a layout — it calls a tool that returns the same structured JSON a script would get from /v1/latest. Our AI and agent access policy states this directly: citation from rendered pages is welcome, but programmatic access should go through the API or MCP, not scraping.

One tool per endpoint

Our MCP server mirrors the REST API one-to-one: one MCP tool per endpoint, same parameters, same response shape. An agent that can call getLatest over MCP is calling the identical logic behind GET /v1/latest?country=germany&indicator=gdp — see getting GDP for any country in one request for that call worked through directly against the REST API, which is useful context even if your agent will ultimately call it over MCP instead.

What's covered

Coverage over MCP is identical to the REST API: 50 countries across 20 economic indicators — GDP, inflation, unemployment, policy rates, government finances, and more — plus bilateral trade data by product and partner for 41 of those countries, like Germany's export mix. Nothing is held back for one access method over the other.

A typical agent workflow

A research agent tasked with "compare inflation across the G20 this quarter" can call the MCP equivalent of /v1/latest?country=g20&indicator=inflation-rate in one tool call, get back 20 source-attributed values, and reason over structured numbers instead of trying to extract them from prose. That same query, run manually, is the inflation indicator hub sorted by country — useful for a person, but the point of MCP is that an agent doesn't have to read it that way.

Getting connected

MCP access is free, same as the REST API — no separate license, no sales conversation. Read the AI and agent access policy linked above in full for what agents may and may not do with this data — citation, bulk extraction, training use — before you point a crawler or an agent at the site.

FAQ

Common questions

Related articles

Connect AI to EconomicsAPI

Trade flows and economic data for AI agents