FXMacroData

What it is

A macroeconomic data API built for currency traders: policy rates, inflation, employment, GDP and about a hundred other indicators for 22 currencies, where every row says when the number became public. The coverage catalogue generated on 07.10.2026 lists 104 indicators and 484 currency-indicator pairs, all taken from official publishers — the Fed, the BLS, the ECB, the ONS, the RBA and their equivalents. The vendor says it passes through nothing from commercial data vendors.

Around the releases sit a forward release calendar, CFTC Commitments of Traders positioning, official FX reference rates, sovereign yield curves, gold and other commodities, and central-bank press releases, all in one schema across currencies.

The release timestamp is the reason to buy it. Each record carries an announcement_datetime, and revisions are kept as separate entries rather than written over the first print, so a backtest can ask what the market knew at 08:30 on release day rather than what the series says now.

Pricing

The free tier is keyless and US-only: the trailing 90 days of US releases, each held back 15 minutes, 100 requests a day, plus the US calendar and recent US COT. On 07.10.2026 an unauthenticated call for US CPI stated both limits in its response, and a call for the euro-area policy rate returned 401.

The two paid plans carry the same data. Individual is $50 a month or $500 a year, for one person's own research or trading; Business is $250 a month or $2,500 a year, licensed to a company for up to ten people. Both allow 1,000,000 requests a month, 300 a minute, 5,000 an hour and 20 concurrent. Both start with a 14-day trial that needs a card, convert automatically, and are non-refundable once billed. Showing the data to other people is a $10-a-month add-on per audience block of 100 users or 1,000 public followers, for attributed display only. Enterprise is unpriced.

Data & coverage

History depth is uneven, and the catalogue prints a start date per pair — check the one you need before subscribing. US CPI starts in January 1914, Canadian CPI in 1915, Swiss CPI in 1922. US nonfarm payrolls start in December 1996, and Australian CPI only on 30.04.2025. Of the 484 pairs, 21 show no start date at all.

The FX rates are central-bank reference rates and fixings, such as Brazil's PTAX and the Central Bank of Iceland's 16:00 rate, not a dealer quote. Releases arrive in real time on paid plans over REST polling, a server-sent-events stream and webhooks.

Forecasts come in classes, and the class matters. Official ones are imported from publisher archives — Philadelphia Fed SPF vintages since 2003, Cleveland Fed CPI nowcasts from the August 2013 reference period. Where no official forecast exists, the API returns FXMacroData's own model output, archived only from the August 2026 reference period.

Integrations

REST with the key in an X-API-Key header, described by an OpenAPI schema of 29 paths. Webhooks are HMAC-SHA256-signed; they and the event stream are paid-only and both answered an unauthenticated request with 401. A hosted MCP server answered an unauthenticated initialize on 07.10.2026 as version 1.29.1, serving the delayed US data without a key and protected tools over OAuth or a bearer key.

The Python client fxmacrodata on PyPI is the vendor's own, MIT-licensed, 1.2.2 uploaded 06.10.2026. Excel works through Power Query's From Web import and Google Sheets through an Apps Script recipe; neither is an add-in. The founder has also contributed a Meltano tap, a DuckDB community extension and an opam package to other projects.

Limitations

• It is one person's business, and a young one. ABN Lookup lists the business as an individual sole trader, and the vendor dates its founding to November 2025. Support on both self-serve plans is founder email, and uptime and response targets are goals, not contractual credits.
• Not every timestamp was observed. Where official calendars do not reach back, release times are derived from the series' median publication lag and can land a few days off. In the vendor's own example, 997 of 1,349 US CPI rows are derived. One derived row turns the whole-response point_in_time_safe flag false, so a strict backtest filters row by row.
• The MCP tools go past data. The MCP tool reference, read on 9 October 2026, documents fx_trade_setup_task as building a trader-facing setup for a pair you name — a directional bias, ranked catalysts, an execution window and invalidation notes. A bias on a currency pair is a call, not a number. Other task tools rank releases by risk, classify macro regimes, run carry and momentum backtests and stress a portfolio; the API's own forecasts are one model's view, not a consensus.
• Consensus is thin. Many releases have no official pre-release consensus, and no plan adds one.
• The terms name no jurisdiction — governed by the principles of international commercial law, arbitration at an unnamed institution. Internal use only without the add-on, and retained data may not feed a competing product.

Alternatives

FRED is free, holds 851,100 series and keeps point-in-time vintages through ALFRED — the first stop for US-only work, though its vintages are dated to the day rather than the minute and it has no COT. Trading Economics covers far more countries, but its calendar API is in no self-serve plan, where FXMacroData's is in the $50 one. DBnomics is the free aggregator over official statistics if you need values rather than release times.

The rest of the field is in macro and economic data.