Sports MCP Servers : Odds & Betting APIs for AI Agents

We checked all 33 providers in our directory for Model Context Protocol support. Here is who actually ships a server, who relies on the community, and who only serves docs.

Last verified: August 2026

What Is a Sports MCP Server?

A sports MCP server exposes a sports odds or data API as tools an AI agent can call directly. MCP is the Model Context Protocol, the open standard that lets tools like Claude, Cursor, and other AI coding agents talk to external services without custom integration code. Add the server to your client, hand it an API key, and your agent can pull live odds, schedules, and scores mid-conversation.

Why does this matter for betting apps specifically? Because odds data is awkward to integrate blind. Market names differ between providers, bookmaker coverage varies by sport, and payload shapes surprise you. When your coding agent can query the real API while it writes your integration, it stops guessing. It sees actual market keys, actual bookmaker lists, actual timestamps, and writes code that matches reality on the first pass.

The catch: "MCP support" on a marketing page can mean three very different things. We verified every provider in our directory and split them honestly into official full servers (live data as tools), docs-only servers (documentation search, no data), and community-built servers (third-party, use with care). No other sports API list makes that distinction. Most get it wrong by counting all three as a green tick.

MCP Support Across 33 Sports APIs: The Verified List

We researched every provider in the SportsAPIs.dev directory in August 2026: their docs, their GitHub orgs, npm, and the major MCP registries. The result, in one table. The MCP pill on our directory cards stays in sync with this research.

Disclosure: Some links on this page are affiliate links. They cost you nothing extra and help keep this directory free.

Provider MCP Status What You Get
SportsGamesOdds ✅ Official full server First-party npm package (sports-odds-api-mcp, Apache-2.0) with live odds tools
Odds-API.io ✅ Official full server First-party npm package (odds-api-mcp-server) with a setup guide in their docs
Odds-API.net ✅ Official full server First-party @odds-api/mcp package, agent-oriented with a mock mode for keyless testing
SharpAPI ✅ Official full server MCP listed in their docs alongside the TypeScript and Python SDKs
PropLine ✅ Official full server First-party server in their GitHub org, shipped with Python, Node, and CLI SDKs
Live Tennis API ✅ Official full server First-party read-only server, listed in the official MCP Registry
Sportradar ❓ Official, docs-only Official server (launched Feb 2026) for documentation and endpoint discovery, not live data
TheRundown ❓ Official, docs-only Official /mcp endpoint for searching endpoints and parameters; live data stays REST-only
The-Odds-API ❓ Community-built Several third-party servers on GitHub and MCP registries; nothing first-party
SportsData.io ❓ Community-built Third-party servers and auto-generated wrappers; nothing first-party
iSports API ❓ Community-built Covered by a community sports-hub aggregator server only
The other 22 providers ❌ None known No official or community MCP server we could verify as of August 2026

Six real servers out of 33 providers. That number will grow fast, and we update the directory as it does. If you are a provider and we missed your server, tell us and we will verify and list it.

The Six Official Sports Odds MCP Servers

SportsGamesOdds

MCP Server: First-party npm package sports-odds-api-mcp, open source under Apache-2.0, maintained in their own GitHub org with a dedicated AI-coding guide in the docs.

The API behind it: Strong US sportsbook coverage including Bet365 and Circa, WebSocket streaming, bet settlement, and a generous free tier. A featured partner that is active in our Discord.

Best for: Agent-assisted development against US books, with a free tier to test on before paying.

Odds-API.io

MCP Server: First-party npm package odds-api-mcp-server with a Claude setup walkthrough published in their docs.

The API behind it: 265+ bookmakers across 34 sports, sub-150ms responses, free tier at 100 requests/hour with no card required. Readers get 30% off for 3 months through our link.

Best for: Global bookmaker coverage in an agent workflow, including hard-to-find books like Bet365 and SBOBET.

Odds-API.net

MCP Server: First-party @odds-api/mcp package. The standout detail is a mock mode: your agent can explore realistic response shapes before you even create an API key.

The API behind it: Pregame and closing odds with Bet365 and Pinnacle coverage, settlements, deeplinks, and stats.

Best for: Trying agent-driven integration with zero signup friction, thanks to the mock mode.

SharpAPI

MCP Server: Official server documented alongside their TypeScript and Python SDKs, treated as a first-class client rather than an afterthought.

The API behind it: Pregame and live odds with Bet365 and Pinnacle, futures, deeplinks, stats, and WebSocket streaming. Free tier available.

Best for: Sharp-book coverage (Pinnacle plus Bet365) inside an agent workflow.

PropLine

MCP Server: First-party server in their GitHub org, part of an official SDK lineup covering Python, Node, MCP, and a CLI.

The API behind it: Props-focused odds with Pinnacle coverage, closing odds, settlement, stats, and WebSocket. Free tier available.

Best for: Player-props tooling where your agent needs to query prop markets directly. See our player props API guide for the wider landscape.

Live Tennis API

MCP Server: First-party read-only server, listed in the official MCP Registry. The cleanest registry presence of any provider here.

The API behind it: Tennis-only depth: ATP, WTA, Challenger, and ITF with point-level live scores, match-winner pricing, and a win-probability model over REST and WebSocket. Plans start at $9.99/mo and readers get 10% off through our link.

Best for: Tennis models and live-betting tools built with agent assistance.

Compare these providers on everything else

MCP is one pill among fifteen. Filter the full directory by free tier, WebSocket, Pinnacle coverage, and MCP support side by side.

Browse the API Directory →

How to Connect a Sports Odds MCP Server

Most official servers ship as npm packages, so setup is short. In Claude Code, one command registers a server for your project:

claude mcp add sports-odds -- npx -y sports-odds-api-mcp

In Cursor or Claude Desktop, add a block to your MCP config file instead:

{
  "mcpServers": {
    "sports-odds": {
      "command": "npx",
      "args": ["-y", "sports-odds-api-mcp"],
      "env": {
        "API_KEY": "your-api-key"
      }
    }
  }
}

Two things to check in the provider's docs before you copy-paste: the exact package name, and the environment variable the server expects for your key. Both vary by provider. Your API key works the same as it does over REST, and every request your agent makes counts against your plan limits, so keep an eye on usage while an agent is exploring.

New to odds APIs entirely? Start with our getting started guide, then come back once you have a key.

Docs-Only and Community Servers: Read the Fine Print

Docs-only servers are official but limited. Sportradar and TheRundown both run MCP servers that help an agent search their documentation, find the right endpoint, and understand parameters. Genuinely useful during integration, but no live data flows through them. Your agent still writes REST calls, and you still test them yourself. If a provider says "we have MCP", ask which kind before you count on it.

Community servers fill the gap for popular APIs whose owners have not shipped one. The-Odds-API has several third-party servers precisely because it is so widely used. Before you rely on one: read the source, since it will handle your API key. Check the last commit date, because an abandoned wrapper breaks silently when the API evolves. And prefer read-only servers with no scopes beyond what you need.

Our position: a community server is fine for prototyping, but for anything you ship, pick a provider that owns its MCP server. The six above treat agent access as a product surface, which means it gets maintained when the API changes.

One more honest caveat: this space moves monthly. A provider marked ❌ today may ship a server next quarter. We re-verify when providers tell us and on our regular directory reviews, and the directory pills reflect the latest check. The 1,300+ developers in our Discord tend to surface new servers within days of launch, several providers are in there answering questions themselves.

Frequently Asked Questions

A Model Context Protocol server that exposes a sports data or odds API as tools an AI agent can call directly. Instead of writing integration code first, you add the server to Claude, Cursor, or another MCP-capable client, and the agent can fetch odds, schedules, and scores itself while it works.
As of August 2026: SportsGamesOdds, Odds-API.io, Odds-API.net, SharpAPI, PropLine, and Live Tennis API ship official full servers. Sportradar and TheRundown run official docs-only servers. Community servers cover The-Odds-API and SportsData.io. Filter the directory by MCP Server →
A full server returns live data: ask for current NFL spreads and real prices come back. A docs-only server searches the provider's documentation and endpoints but returns no data; you still write and authenticate REST calls yourself. Both are useful, but they are not the same feature.
Treat them like any third-party dependency that handles a credential. Read the source before giving one your API key, check the last commit date, and prefer read-only servers. Fine for prototyping; for production, prefer a provider that owns its server.
Almost always. The server passes your key through to the API, usually via an environment variable in the config, and agent requests count against your plan limits exactly like REST calls. Watch your usage while an agent explores; agents are curious. Several providers offer free tiers that are perfect for this.
No. Every server listed here is read-only data: odds, schedules, scores, stats. Odds APIs supply the data your application uses; they do not place bets, over MCP or otherwise.
Not on its own. Coverage, latency, and pricing still come first; see our sports odds API guide for how to weigh them. But between two comparable providers, take the one with an official MCP server. It saves integration time today and signals ongoing investment in developer experience.
Anything you build with the API, faster: odds comparison tools, live betting dashboards, arbitrage scanners, model backtests against historical odds. Plus a new category: research agents that answer questions like "which book moved first on this line" directly in your AI client.

Building With AI Agents?

Join our Discord community of 1,300+ developers building betting tools. Ask which MCP servers people actually use, and get answers from developers and providers in the same channel.