The Free COT API

Every endpoint is a static file on a CDN: no keys, no login, no rate limits, CORS open. Data updates within minutes of each Friday CFTC release.

Endpoints

Market slugs are listed in llms.txt and linked from every page under /cot/. /api/v1/cot/latest.json serves the same bytes as /api/v1/latest.json, which is the canonical path.

Which report family a market has

The CFTC does not publish all three report families for every market, so .disagg.csv and .tff.csv exist only where the underlying report does. The split follows the CFTC's own: the disaggregated report covers physical commodities (grains, energy, metals, softs, livestock), and Traders in Financial Futures covers financials (equity indices, interest rates, currencies). So sp500-emini.disagg.csv does not exist and sp500-emini.tff.csv does; for corn it is the other way round. The legacy .csv and .json exist for every market.

You never have to guess. Every market in /api/v1/latest.json carries a datasets object listing exactly the URLs that exist for it:

{"markets": {"sp500-emini": {"datasets": {
    "csv":      "https://futuresbench.com/data/cot/sp500-emini.csv",
    "json":     "https://futuresbench.com/data/cot/sp500-emini.json",
    "tff_csv":  "https://futuresbench.com/data/cot/sp500-emini.tff.csv"}}}}

Read the keys rather than building the path by hand and a new family appearing for a market picks itself up.

The CFTC's own datasets (Socrata IDs and flat files)

If you would rather pull straight from the source, these are the six datasets our pipeline ingests every Friday. The CFTC publishes each report twice: as a Socrata dataset on publicreporting.cftc.gov (queryable, full history) and as a current-week flat file under cftc.gov/dea/newcot/.

One week of one report, as CSV (the $where date is the Tuesday the positions are as of):

https://publicreporting.cftc.gov/resource/gpe5-46if.csv?$where=report_date_as_yyyy_mm_dd='2026-09-22'&$limit=5000

The whole dataset as one download is https://publicreporting.cftc.gov/api/views/gpe5-46if/rows.csv?accessType=DOWNLOAD (swap in any ID above). Things we learned running this every week:

• Futures-only vs combined is the easy mistake. kh3c-gbw2 and yw9f-hn96 look like the disaggregated and TFF datasets you want, but they are the combined (futures plus options) variants. Mixing the two gives you positions that do not reconcile week to week.
• The flat files sit behind a CDN cache. Right after a release the newcot file can still serve the previous week for a while, and cache-busting does not help. Socrata is not CDN-fronted, so it is the authoritative fallback.
• The six reports do not always land together. On 2026-09-25 the legacy report was live at release time while the disaggregated and TFF reports were still on the prior week in both sources. Check each dataset's latest date rather than assuming one release means all six.
• cftc.gov rejects some default HTTP clients. A request with Python's default urllib user agent gets HTTP 403; send a descriptive User-Agent.

Stability promise

The v1 CSV and JSON schemas are additive-only: existing columns and keys never change meaning, get renamed, or disappear. New columns and keys may appear. A breaking change would ship under a new path (/api/v2/), never in place. Null values are empty CSV cells / JSON null - never zero.

Examples

curl

curl -s https://futuresbench.com/api/v1/latest.json | head -c 400

Python / pandas

import pandas as pd
gold = pd.read_csv("https://futuresbench.com/data/cot/gold.csv",
                   parse_dates=["report_date"])
print(gold.tail())

Google Sheets

=IMPORTDATA("https://futuresbench.com/data/cot/all-latest.csv")

JavaScript

const r = await fetch("https://futuresbench.com/api/v1/latest.json");
const { report_date, markets } = await r.json();
console.log(report_date, markets.gold.net_noncommercial);

Fair use

There are no keys and no enforced limits - the endpoints are CDN-cached static files and normal use costs us nearly nothing. If you're polling for the weekly release, once a minute from 3:30 p.m. ET on Fridays is plenty (the data is weekly). For bulk backfills, each market is one file - please don't re-download unchanged history on a schedule.

Questions or a use case that needs more: data@futuresbench.com.

Source & licensing

Underlying data: CFTC Commitments of Traders reports, U.S. federal public domain. Derived metrics (net positions, COT index, z-scores) are computed as documented in the methodology. A link back to futuresbench.com when you publish something built on this data is appreciated.