BLS Public Data API

What it is

The BLS Public Data API is the Bureau of Labor Statistics' own endpoint for the series it computes: the Consumer Price Index, the establishment and household employment surveys, producer prices, the Employment Cost Index, JOLTS, productivity and import/export price indexes. Series ID formats are documented for around twenty-five programmes, each with its own prefix — CU for CPI-U, CE for national employment, WP for PPI commodity data, JT for JOLTS.

A request is a JSON POST to api.bls.gov/publicAPI/v2/timeseries/data/ carrying a list of series IDs and a year range; a single series can also be pulled with a plain GET. Responses are JSON, with each observation carrying its year, period, value and a footnote array — which is where the preliminary flags live.

Most readers meet these series second-hand, through FRED or an aggregator. The reason to come to the source is the footnotes: this is the number as the agency published it, with the agency's own caveat attached.

Pricing

Free in both versions; registration buys headroom, not data. The step from v1 to v2 is 25 queries a day to 500, 25 series per query to 50 and 10 years of history to 20, plus calculations, annual averages and catalog metadata. Both share a ceiling of 50 requests per 10 seconds, and the Terms of Service warn that trying to circumvent it can get your access blocked. Registration takes an email address, an organisation name and a CAPTCHA, and the key must be renewed annually.

Worth knowing before you plan around the versions: the v2 path answers without a key. A keyless POST on 19 September 2026 returned the data alongside a message array saying the catalog and the calculations had been disabled for the request. The key gates the extras and the quota, not the numbers.

Data & coverage

United States only, and nothing priced. Checked on 19 September 2026, the August 2026 readings were already there: CPI-U all items (CUUR0000SA0) at 334.980, total nonfarm payrolls (CES0000000001) at 159,075 thousand and PPI final demand (WPUFD4) at 157.604 — the last two flagged preliminary. A two-to-three-week lag on the monthly headline series is the shape to plan against. Monthly is the fastest frequency in the building; much of the catalogue is quarterly or annual.

Integrations

BLS ships no client library, no spreadsheet add-in and no MCP server. The Python packages people use are third-party and both are dormant: bls on PyPI is Oliver Sherouse's, last released in June 2018, and bls-api is Tien Nguyen's, last released in September 2022. Neither is named in the BLS documentation.

Spreadsheet output is a URL, not a parameter: post the same JSON body to api.bls.gov/publicAPI/v2/timeseries/data.xlsx, or append .xlsx to a single-series GET, and the response is a real workbook with a Content-Disposition filename. An output_type field in the body does nothing — a request carrying one came back as ordinary JSON on 19 September 2026.

For bulk there is a flat-file server at download.bls.gov/pub/time.series/, one directory per survey, tab-delimited, with timestamps that track the release calendar. The BLS web hosts are unfriendly to anonymous automation — on 19 September 2026 a plain curl call to www.bls.gov was refused with 403 while the same request with a User-Agent naming a contact address was served. The API host itself does not do this.

Limitations

• This is not market data. No instrument, no price, no volume, nothing intraday — it is the macro input to a model.
• No point-in-time access. The request signature has no vintage parameter, so a backtest reading today's CPI is reading numbers nobody had at the time. ALFRED, on the FRED side, is where that problem is solved.
• Revisions differ per programme and live in the footnotes. PPI is revised monthly for four months after publication; payrolls arrive preliminary, are revised twice and then benchmarked annually; unadjusted CPI is final on release while the seasonally adjusted version is restated for up to five years as the factors are re-estimated each January.
• 25 queries a day is a development budget, and even 500 is small across a wide panel. This is a pull-on-release API, not something to poll.
• The key expires every year, and the failure mode is an authentication error rather than a notice.
• Public domain, with a sentence attached. BLS asks to be cited and requires you to state that it cannot vouch for data or analyses derived from the data once they have left BLS.gov.

Alternatives

FRED republishes most of these series at 120 requests a minute, with vintages through ALFRED and an Excel add-in, and is the better default unless you want the agency's own footnotes. DBnomics mirrors BLS alongside dozens of other providers under one query shape. EconDB and Trading Economics wrap the same public numbers in a commercial product.