Sports Betting Mcp

sports-betting-mcp

The first MCP server for sports betting. Give any AI agent live access to picks, odds, game analysis, and performance tracking across NBA, NHL, NCAAB, and MLB.

mcp-name: io.github.seang1121/sports-betting-mcp

---

What It Does

Connects any MCP-compatible AI agent to a live sports betting analysis system. Every pick is logged before tip-off and resolved against final scores. Nothing is cherry-picked.

Metric	Value
Sports Covered	NBA, NHL, NCAAB, MLB
Bet Types	Moneyline, Spread, Totals
Pick Source	12-agent consensus model
Total Picks	3,859+ resolved
Platform Users	30
Tools	12 MCP tools
Auth	API key (X-API-Key header)

Results by Sport

Sport	Picks	Wins	Win Rate
NBA	1,267	762	60.1%
NHL	1,148	656	57.1%
NCAAB	1,149	549	47.8%
MLB	283	109	38.5% (launched Apr 2026)

All results are queryable in real-time via the get_win_rate tool. Numbers update after every game.

---

Quick Start

Install

pip install sports-betting-mcp

Add to Your MCP Client

Drop this into your MCP config (Claude Desktop, Cursor, Windsurf, Claude Code, etc.):

{
  "mcpServers": {
    "sports-betting": {
      "command": "sports-betting-mcp",
      "env": {
        "SPORTS_BETTING_API_URL": "https://sportsbettingaianalyzer.com",
        "SPORTS_BETTING_API_KEY": "your_api_key"
      }
    }
  }
}

Environment Variables

Variable	Required	Description
SPORTS_BETTING_API_KEY	Yes	Your API key for authentication
SPORTS_BETTING_API_URL	No	API base URL (defaults to http://localhost:5000)

---

Works With

Any client that supports the Model Context Protocol:

Client	Status
Claude Desktop	Fully supported
Cursor	Fully supported
Windsurf	Fully supported
Claude Code (CLI)	Fully supported
Any MCP Client	Fully supported via stdio transport

---

Available Tools

12 tools. Every call returns structured data that AI agents can reason over, display, or act on.

Picks & Analysis

Tool	What It Does
get_todays_picks	All AI picks with confidence scores and edge breakdowns, filterable by sport
get_top_pick	Single highest-confidence pick of the day
get_pending_picks	Currently unresolved picks that are still in play
get_completed_picks	Recently resolved picks with W/L results -- verify the track record
analyze_game	Full 12-agent analysis on any game: consensus pick + edge breakdown

Odds & Market Data

Tool	What It Does
get_live_odds	Live moneyline, spread, and totals for today's games
get_alerts	Active alerts from the multi-agent system (line moves, injury impacts)

Performance & Stats

Tool	What It Does
get_win_rate	Win rate with breakdown by sport and bet type
get_model_stats	Model performance: total picks, last-20 win rate, confidence tier breakdown
get_leaderboard	User rankings by win rate
get_system_status	Health check -- uptime, database, scheduler status

---

How the Analysis Works

Each game runs through a multi-agent pipeline:

1. 12 specialized agents evaluate the game independently -- covering momentum, matchups, injuries, rest, travel, public betting percentages, sharp money indicators, historical trends, and more.
2. A consensus engine synthesizes all 12 opinions into a single pick with a confidence score.
3. Edge calculation compares the model's implied probability against the current market line.
4. Picks are logged before tip-off and resolved against final scores. No retroactive edits.

The confidence score and edge breakdown are included in every pick response, so your AI agent can filter, rank, or explain the reasoning behind any recommendation.

---

Full API — 22 Endpoints

The API has two tiers: Free (any API key) and Pro (approved access).

Free Tier (available to all API key holders)

Endpoint	Method	What It Does
/xk/p	GET	Today's AI picks with confidence scores
/xk/o/{sport}	GET	Live odds from FanDuel/BetMGM
/xk/w	GET	Win rate breakdown by sport and bet type
/xk/i	GET	Active injury report from Covers.com
/xk/m	GET	Significant line movement (sharp money signals)
/xk/games	GET	Today's full schedule with odds snapshot and AI pick flags
/xk/stats	GET	Model performance by sport, bet type, confidence tier
/xk/lb	GET	Leaderboard — AI vs human win rates
/xk/news	GET	Sports news feed from RSS sources
/xk/q	GET	Currently pending (unresolved) picks

Pro Tier (requires approved access)

Endpoint	Method	What It Does
/xk/full	GET	Full structured picks — complete edge breakdowns, positive/negative edges, reasoning, scorecard, opposing side with odds, line movement, quality flags
/xk/analyze	POST	On-demand 12-agent analysis for any game — send team names, get full consensus
/xk/results	GET	Resolved picks with W/L, actual scores, profit, top contributing edges
/xk/clv	GET	Closing Line Value analysis — did the line move in our favor?
/xk/search	GET	Search all picks by team name with W/L record
/xk/log	POST	Log a pick (supports flip flag, pick source tracking)
/xk/log/batch	POST	Batch log up to 20 picks with per-pick success/duplicate/error status
/xk/history	GET	Your pick history — pending, resolved, win rate
/xk/slip	POST	Generate Nimrod bet slip image
/xk/webhook	POST/GET/DELETE	Register webhooks for pick events
/xk/full	GET	Includes fade/flip data for every pick — opposing side, line, odds

Pro Pick Payload

Every pick from /xk/full includes:

{
  "ai_pick": "Lakers -5.5",
  "ai_verdict": "STRONG BET",
  "ai_probability": "65%",
  "ai_reasoning": "Lakers defense ranks top 5, opponent on B2B...",
  "edges": [13 individual edge factors],
  "top_edges": [top 5 positive],
  "negative_edges": [top 3 negative],
  "scorecard": [3 model scores],
  "fade_pick": "Celtics +5.5",
  "fade_team": "Celtics",
  "fade_odds": "+105",
  "fade_note": "Betting AGAINST the AI — take Celtics +5.5",
  "line_moved": true,
  "flags": ["line_moved"],
  "data_quality": "good"
}

Pick Source Tracking (for learning)

When logging picks, include pick_source to track decision types:

• model_agree — following the AI's pick
• flip — fading/betting against the AI
• manual_override — custom pick

This enables win rate comparison between model-agree vs flip picks over time.

---

API Authentication

Authenticate with X-API-Key header. Get a free key at sportsbettingaianalyzer.com. Pro access requires admin approval.

---

Tech Stack

Component	Technology
Runtime	Python 3.10+
Protocol	MCP (Model Context Protocol)
Transport	stdio
Build	Hatchling
Distribution	PyPI (sports-betting-mcp)
Backend	Flask + SQLite
Analysis	12-agent consensus pipeline
Sports	NBA, NHL, NCAAB, MLB

---

Related Projects

• Mortgage Rates MCP Server -- 17 lenders, 13 tools, real-time mortgage rate intelligence
• March Madness Bracket Predictor -- 5-model ensemble, 14-year backtest, 77% accuracy, correctly predicted 2026 champion
• Agent Command Center -- Dashboard that auto-discovers MCP servers, agents, hooks, cron jobs, and repos

License

MIT