Introduction

sportsclaw is an open-source engine for building AI that understands live sports.

Most AI assistants are guessing when you ask about sports — they don't know last night's score, tonight's odds, or whether the game just went to overtime. sportsclaw fixes that. It connects your AI model to real, live sports data and gives it the tools to follow a game in real time, answer with the actual numbers, and reach out the moment something happens.

It's not a coding copilot and it's not a generic chatbot framework. It's a sports-native engine: the data, the live game events, and the real-time plumbing are already wired in, so you spend your time building the experience — not stitching together feeds.

What you can build

• Sports chat bots for Discord and Telegram that answer with real scores, standings, and odds — and ping followers when their team plays.
• Broadcast and studio tools that pull live game state and generate graphics on demand.
• Odds and market trackers that watch ESPN, Kalshi, and Polymarket and surface the edge.
• Anything agentic — sportsclaw is a TypeScript engine you can embed, extend, and deploy your own way.

Who it's for

Developers and technical teams building for sports fans, broadcasters, sportsbooks, and teams & leagues. If your product needs an AI that genuinely keeps up with live sports, this is the foundation.

What's in the box

• 14 sports — NFL, NBA, WNBA, MLB, NHL, soccer, F1, tennis, cricket, golf, volleyball, college football, college basketball, and track. No API keys needed for the data.
• Live odds & prediction markets — ESPN, Kalshi, and Polymarket, with betting math built in.
• Real-time game events — the engine watches games and knows when the score, lead, or status changes.
• Bots, images, and vision — Discord/Telegram integrations, image generation, and the ability to read images you send it.

All coverage is powered by sports-skills, the open data layer the installer sets up for you. When you need licensed, real-time feeds, there's a premium tier — see Machina.

Bring your own model — Anthropic (Claude), OpenAI, or Google (Gemini).

Ready in about a minute

Head to the Quickstart to install sportsclaw and run your first query.