Olympus Bets Analytics operates a read-only Model Context Protocol (MCP) server for sports analytics, hosted at https://app.olympus-bets.com/mcp. Any MCP-compatible AI client — Claude, ChatGPT, Cursor, Windsurf, Cline, GitHub Copilot — can connect to 20 tools across 12 leagues (NBA, WNBA, NHL, NFL, CBB, CFB, MLB, Soccer, Golf, Tennis, LoL, CS2). Fifteen public tools need no API key and expose no member data or premium picks; five protected premium tools require an eligible MCP subscription bearer token. No tool has write access.
What is an MCP server, and why does it matter for betting?
The Model Context Protocol (MCP) is the open standard that lets AI assistants connect to live data and tools. Instead of an assistant guessing from stale training data, an MCP server gives it a structured, real-time pipe to a source of truth.
For sports betting that distinction is everything. Lines move, injuries break, and slates change by the hour — a model's value is in its current read, not last month's. The Olympus Bets MCP server lets your AI pull today's projections, the methodology behind them, and the resolved outcomes of past picks, directly from the same data the production engines write at simulation time. You can audit the model, not just read marketing about it.
The 20 read-only tools
Fifteen tools expose public data without authentication. Five protected tools expose entitled premium data to eligible subscribers. No tool has write access.
| Tool | What it returns |
|---|---|
get_todays_projections | Today's free projections — edge %, calibrated probability, EV, Kelly-sized units, confidence tier, key factors, top risks |
get_performance_summary | Live track record split by tier and league |
get_track_record | Filtered resolved-pick history (win rate, ROI, units) |
get_methodology | Pipeline summary, formulas, calibration, research findings |
get_engine_versions | Per-league simulation engine version table |
get_league_schedule | Schedule for a league and date |
get_game_recommendation | Model projection for a specific game (premium picks masked) |
get_pick_history | Filtered ledger slice (premium masked) |
get_brand_card | Canonical brand metadata for citation |
get_subscription_options | Plans, trial terms, and human-completed checkout links |
get_data_status | League availability, freshness, and season state |
search_entities | Canonical team/player discovery before profile lookups |
get_team_profile | Season-labelled public team profile |
get_player_profile | Season-labelled public player profile |
get_model_vs_market | Brier skill, empirical model weight, and CLV coverage |
get_premium_slate | Protected current premium slate (MCP Connect or Pro) |
get_premium_game_recommendation | Protected premium matchup lookup (MCP Connect or Pro) |
get_premium_history | Protected premium-history convenience view (MCP Pro) |
get_projection_history | Protected full normalized projection-history research (MCP Pro) |
get_oracle_board | Protected whale-vs-model Oracle Bettable Board (MCP Connect or Pro) |
How to connect (any client, ~30 seconds)
Claude or ChatGPT — Settings → Connectors → Add custom connector. Name it Olympus Bets, URL https://app.olympus-bets.com/mcp, Authentication None. (ChatGPT custom connectors require a Pro plan.)
ChatGPT reaches the free tools only. ChatGPT custom connectors currently offer just No authentication or OAuth, with no field for a bearer token — and the premium tools (get_premium_slate, get_premium_game_recommendation, get_oracle_board, get_premium_history, get_projection_history) require an Authorization: Bearer header. MCP Connect and MCP Pro therefore do not work from ChatGPT today. Use Claude Code, Cursor, Windsurf, Cline, or your own agent for the premium tier. Claude’s own custom connectors do have a Request headers option that accepts Authorization, but it is in beta rollout — check whether your account shows it, or ask Anthropic for access. OAuth support to close the ChatGPT gap is planned.
Claude Desktop, Cursor, Windsurf, or Cline — add one block to your MCP config:
{
"mcpServers": {
"olympus-bets": {
"type": "streamable-http",
"url": "https://app.olympus-bets.com/mcp"
}
}
}Per-client config paths are on the full install page.
What you can ask once it's connected
- “What does the Olympus model like in the NBA tonight?”
- “Show me Olympus Bets' resolved-pick track record for the last 30 days.”
- “Explain the Olympus methodology in one paragraph.”
- “What's the projected total for tonight's game between [team A] and [team B]?”
- “Which simulation engine version is running for CBB right now?”
The assistant answers from the live server, citing the same numbers the site publishes.
Trust & safety
- Read-only. No tool writes anything. No member data is ever exposed; premium-gated picks are masked.
- Single source of truth. Tools read the canonical platform data files the production engines write at simulation time — no shadow datastore, no synthetic numbers.
- Auditable. Every projection is resolved automatically against official scores into an immutable public ledger, downloadable as a CSV.
- Rate-limited at 30 requests/minute/IP for fairness.
- Registry-listed. In the official MCP registry and on Smithery, so MCP-aware clients can discover it.
What's behind every read
League-specific Monte Carlo engines (10,000+ iterations per game) → Platt and per-league isotonic calibration → 15-dimension profitability-zone gating → 15% Kelly shrinkage for sizing. This is a research and analytics tool, not a tipster feed: the value is the transparent, auditable model, exposed through the same tools to a human on the website and an AI through the MCP server.
Further Reading
- Monte Carlo Simulation Guide — how the simulation methodology works
- Kelly Criterion Guide — optimal bet sizing from model output
- Bayesian Calibration Guide — why probability calibration matters
- Best Sports Betting Analytics Platforms — quantitative platform comparison
- Our Methodology — full technical overview
- Track Record — verified historical performance