DunkAPI Review 2026

DunkAPI is a basketball data API for the NBA, the WNBA, men's and women's NCAA Division I and Australia's NBL. It has 20 endpoints for box scores, play-by-play with shot locations, standings, team ratings, injuries and sportsbook odds. You call each one with a POST to mcp.dunkapi.com/v1/{tool} and a JSON body. The same tools also run as an MCP (Model Context Protocol) server, so an AI app can call them directly.

It is a stats and odds-history product, not a live feed. If you need in-play scores, pair it with a live source from the sports data API list.

NBA play-by-play and shot locations run from the 2003-04 season to the present, at about 450 plays per game. Each play carries the clock, score, players involved, shot type, zone and an expected-points value. Box scores cover the same span, with gaps in the 2012-13 to 2017-18 seasons that the docs list (checked November 2026).

Other data types:

• Standings, schedules and team ratings: an Elo rating before and after every game.
• Sportsbook lines: moneyline, spread and total by book, with opening and closing lines from July 2020.
• Player props: closing lines by book from May 2023, plus line movement and over/under history.
• Injury reports: captured daily for the NBA and WNBA.

Pass league as wnba, ncaam, ncaaw or nbl to switch league. Depth varies by league and dataset, and the data atlas on the DunkAPI site shows what each one holds. For older line history, see our historical odds guide.

DunkAPI bills in credits, and every plan reaches all 20 endpoints. A team list costs 1 credit, standings cost 2, a box score costs 5 and play-by-play costs 10. The pricing page also prices a full season of each dataset, so you can budget before you build.

The provider is plain about timing. Odds and props are pre-game only, and every row carries a captured_at time. Box scores and play-by-play refresh once a day at 10:00 UTC. A public page spells out what the terms allow: you can store the data, train models and build products on it, but you cannot resell it as a raw feed.

The free plan gives 500 credits a month with no card, which covers 250 standings calls or 50 play-by-play calls (checked November 2026). Starter costs $19 a month for 10,000 credits. Pro costs $49 for 30,000 credits, and Scale costs $149 for 125,000.

You can also top up a wallet from $5, at 400 credits per dollar, and those credits do not expire while your account is open. Enterprise accounts get custom rate limits and volume pricing on request. For other no-cost options, browse the free sports APIs we track.

You sign up at the DunkAPI sign-up page. The free plan gives you 500 credits a month and needs no card (checked November 2026). DunkAPI verifies your email first, then you create a key from the dashboard.

Send the key in an x-api-key header, or as Authorization: Bearer YOUR_API_KEY. The MCP endpoint also takes a ?key= query parameter. DunkAPI tops a free account back up to 500 credits within about 3 days after each monthly anniversary of your signup, if the balance is lower. A paid plan starts at $19 a month.

This sample posts to /v1/get_standings on https://mcp.dunkapi.com with the key in x-api-key (from the docs, checked November 2026). The call costs 2 credits.

import os

import requests

resp = requests.post(
    "https://mcp.dunkapi.com/v1/get_standings",
    headers={"x-api-key": os.environ["DUNKAPI_API_KEY"]},
    json={"season": 2026},
    timeout=10,
)
resp.raise_for_status()

for team in resp.json()["standings"][:5]:
    print(team["abbrev"], team["wins"], team["losses"])

The response holds the season and a standings list. Each team row has fields such as abbrev, display_name, wins and losses. Every DunkAPI endpoint is a POST with a JSON body, so the pattern stays the same for the other 19.

The rate limit depends on your balance, not your plan. With 100 credits or more you get 60 requests a minute. Under 100 credits you get 10 a minute (checked November 2026). Enterprise accounts can ask for custom limits. A 429 means you hit the limit, a 402 means you ran out of credits, and a 500 refunds its credits. REST responses carry X-Credits-Used and X-Credits-Remaining headers, and an x-idempotency-key header makes retries safe.

The docs list every endpoint with parameters and a sample response. DunkAPI also publishes an OpenAPI 3.1 spec, an llms.txt file and a tool catalogue. The official Python client is dunkapi on PyPI. It has no required dependencies and reads DUNKAPI_API_KEY. A status page shows measured uptime and data freshness. Support is support@dunkapi.com.

DunkAPI fits three groups well:

• Analysts and modellers who want NBA play-by-play, shot charts and closing lines to backtest against.
• Prop bettors who want line history and hit rates by book.
• AI app builders who want a basketball MCP server with a free monthly allowance.

It covers basketball only. A team that needs live odds or other sports can run it next to a feed from the live odds API list.

DunkAPI does one sport and documents it well. The data atlas, the timing page and the public status page all state what is partial, which makes the numbers easy to trust. A free allowance that refills every month, plus a $19 first paid step, keeps the cost of a trial near zero.

Start with the free key at dunkapi.com, pull one game's play-by-play and compare the shot data with your own source.

Last updated: 5 November 2026 · How providers get listed