Quickstart

The API lives at https://api.betflux.ai. Dataset payloads are per-game Parquet files; the betflux CLI is the fastest way to query them. Start with an active BetFlux account. Access is currently invite-based — join the Discord if you need an invite. After signup, Getting started walks you through the same setup below.

1. Install the CLI.

   uv
   Install uv first (optional)

   Skip this if you already have uv. On macOS or Linux, run:

   curl -LsSf https://astral.sh/uv/install.sh | sh

   Restart your terminal after installation. uv will also install Python if needed. For Windows and other options, see the uv installation guide.

   uv tool install betflux

   Automatic CLI updates are enabled by default on Linux and macOS. See CLI updates for controls and supported installations.

   pip

   Requires Python 3.10 or later.

   pip install betflux

   See Installation.

2. Authenticate the CLI.

   betflux login

   Enter the terminal code in your browser and select Approve access. The CLI saves a key in your system keychain for CLI commands and Python clients. Run betflux keys check to verify access. See Browser login.

3. Install the agent plugin. Choose the command for your agent:

   Codex

   betflux plugin install --codex

   Claude Code

   betflux plugin install --claude

   Your agent’s CLI must already be installed. Start a new agent session after installation. See AI agents for other agents and setup details.

4. Find games. Game ids are readable — LEAGUE_AWAY_HOME_YYYYMMDD — and betflux games is how you discover them. Each row is the full game record: schedule, teams, score, venue, and weather.

   betflux games --league MLB --date-from 2026-07-01 --date-to 2026-07-07

5. Pull a dataset. Fetch one game’s graded market results, filtered and rendered locally:

   betflux get market-results --game MLB_BOS_NYY_20260715 --operator FANDUEL

   Or a date range — the CLI discovers the games, downloads each game’s file, and applies your filters client-side:

   betflux get closing-lines --league MLB \
     --date-from 2026-07-01 --date-to 2026-07-07 --market-type MONEYLINE

No SDK? The files are plain HTTP

Each dataset is served at GET /v1/games/{game_id}/{dataset} as a Parquet file. curl saves it; DuckDB queries it:

curl -s "https://api.betflux.ai/v1/games/MLB_BOS_NYY_20260715/market-results" \
  -H "Authorization: Bearer $BETFLUX_API_KEY" -o results.parquet

duckdb -c "SELECT operator, market_type, side, closing_odds, outcome
           FROM 'results.parquet'
           WHERE market_type = 'MONEYLINE'
             AND closing_odds IS NOT NULL AND closing_ts < game_start"

The closing-field filters keep this example to selections with a pre-game quote, including when reading a historical v2 artifact. In v3, live-only selections remain in market-results with a graded outcome and null closing fields.

DuckDB can also query the authenticated URL directly, without downloading — see Response formats.

Next steps

• Authentication — key format, beta access, and error semantics.
• Datasets — what each dataset contains and how it’s built.
• Rate limits & quotas — per-minute limits and how games per billing period are counted.