"How can an AI agent find the best odds on a game?"

"Three free ways: read agentbets.ai/odds/today/ (every game in the next 24 hours with the best price per side), call GET https://api.agentbets.ai/api/odds/best?event=lakers+celtics (no auth, add \u0026format=text for a plain-English answer), or use the get_best_odds tool on the AgentBets MCP server at api.agentbets.ai/mcp. All three return the best and runner-up book per side, compared across 18+ sportsbooks and refreshed hourly."

"Is this a raw odds data feed?"

"No — deliberately. The endpoint and tool return computed answers only: the best and runner-up price per side for up to 3 matching events, with links back to the full comparison pages on agentbets.ai. Full book-by-book boards are displayed on the site itself. This keeps the service within The Odds API's terms, which allow displaying odds in our product but not redistributing them as raw data."

"What does it cost and what are the limits?"

"Free, no account, no API key. Lookups are limited to 20 per minute and 500 per day per IP — generous for answering user questions, and deliberately useless for reconstructing a bulk feed."

Ask Your AI Who Has the Best Odds Tonight — It Can Actually Check

When someone asks their assistant “who has the best odds on tonight’s Lakers game?”, the assistant needs three things: a current answer, a source it can cite, and a way to get it without signing up for anything. Starting today, AgentBets provides all three.

Three surfaces, one live dataset

The same hourly-refreshed odds data — 18+ sportsbooks across every in-season sport we track — is now reachable however the agent happens to arrive:

A page to read and cite: /odds/today/ lists every game commencing in the next 24 hours with an answer-first sentence per game — “Best moneyline price: Celtics −145 at Bet105; Lakers +132 at BetOnline” — plus the full best-line table and a visible odds-updated timestamp.
An endpoint to call: GET https://api.agentbets.ai/api/odds/best?event=lakers+celtics. No auth, no key. Optional sport and market parameters, and format=text returns the whole answer as one plain-English paragraph an agent can pass straight to its user. The contract lives in /api/openapi.yaml.
An MCP tool: get_best_odds on the AgentBets MCP server, alongside compare_sportsbooks, get_vig_rankings, and recommend_sportsbook.

Try it right now:

curl "https://api.agentbets.ai/api/odds/best?event=chiefs&format=text"

Computed answers, not a feed

Every response returns the best and runner-up book per side — moneyline, spreads, or totals — for at most 3 matching events, with attribution links to the full boards on agentbets.ai. That design is deliberate: The Odds API’s terms allow us to display odds in our own product but not to redistribute them as raw data. So the site displays the full boards, and the programmatic surfaces answer questions about them. Rate limits (20/minute, 500/day per IP) are generous for a conversation and useless for scraping.

Built for how agents actually search

The same principle behind our free listings intake: when someone asks their assistant to do something, the assistant should be able to complete it properly on the first try. Error responses explain exactly what to change (a missed event returns the sports searched and the currently active sports). A machine-readable index of every live odds page ships at /llms-odds.txt, now with today’s games and their best prices inline. And everything is discoverable from /llms.txt, the OpenAPI spec, and the MCP server.

If you’re building an agent that needs to answer odds questions — or you just want your assistant to line-shop for you — point it at any of the three surfaces above. No permission needed; that’s the point.