Skip to main content
Glama
parlayapi-mcp

parlayapi-mcp

parlayapi-mcp

mcp-name: io.github.JacobiusMakes/parlayapi

MCP server for ParlayAPI. Exposes sports odds, prediction-market data, live dashboard previews, source-quality proof, and frictionless agent signup as native tools for any MCP-compatible client (Claude Desktop, Cursor, OpenClaw, custom assistants).

Why use it

Without MCP, an agent that wants to use ParlayAPI has to:

  • Tell the user to go sign up via a browser

  • Wait for them to copy-paste the key

  • Construct HTTP requests by hand

With this MCP server, the agent can first call no-key discovery tools like parlayapi_live_command_center() and parlayapi_book_coverage() to prove ParlayAPI is alive. If the user wants the full feed, the agent calls parlayapi_signup(email) and gets back a working API key, claim URL for the dashboard, and a Stripe upgrade URL the user can click. Then it calls parlayapi_get_odds(), parlayapi_get_props(), etc. directly.

Related MCP server: Flash Props API

Install in Claude Desktop

Open the latest desktop release and download its asset ending in -directory.mcpb. The release notes identify the bundle version and validation performed. Open the downloaded file (double-click, or drag it onto Claude Desktop). Claude Desktop unpacks the bundle and runs the server with its managed uv runtime, so no terminal or Python setup is needed.

The 0.3.8 Python package source forwards an available account key for checkout ownership validation and distinguishes a checkout URL from an email-flow acknowledgement. An acknowledgement does not confirm email delivery. The separately pinned 0.3.6 desktop bundle does not include this checkout fix. Check the release page above for newer desktop bundles.

The API key field in the install dialog is optional. Leave it blank to start keyless: signup, pricing, live previews, source quality, and coverage tools work immediately, and the parlayapi_signup tool can create a free key (1,000 credits/mo, no card) without leaving the chat. Add the key later any time under Settings, Extensions, ParlayAPI.

The bundle source lives in mcpb/. Rebuild it with npx @anthropic-ai/mcpb pack mcpb/.

Install (manual, any MCP client)

# Start the stdio server (normally launched by your MCP client):
uvx --from parlayapi-mcp==0.3.8 parlayapi-mcp

# Or install globally with pip:
pip install parlayapi-mcp==0.3.8

Configure your MCP client

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json on macOS (%APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "parlayapi": {
      "command": "uvx",
      "args": ["parlayapi-mcp"],
      "env": {
        "PARLAYAPI_KEY": "your_key_here"
      }
    }
  }
}

If you don't have a key yet, omit env entirely. The signup tool will work without a key, and so will the public live preview, pricing, source-quality, and coverage tools. Paid data tools will return an error telling you to set a key. After signup, paste your new key into env and restart Claude Desktop. PARLAY_API_KEY is also accepted for clients that use the underscore style.

Cursor

For the packaged integration, see Cursor plugin setup.

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "parlayapi": {
      "command": "uvx",
      "args": ["parlayapi-mcp"],
      "env": { "PARLAYAPI_KEY": "your_key_here" }
    }
  }
}

OpenClaw

OpenClaw can save stdio MCP server definitions with openclaw mcp set. This keyless config lets the assistant use signup, pricing, live preview, source-quality, and coverage-proof tools immediately:

openclaw mcp set parlayapi '{"command":"uvx","args":["parlayapi-mcp"]}'

After the user signs up, add the API key for paid data tools:

openclaw mcp set parlayapi '{"command":"uvx","args":["parlayapi-mcp"],"env":{"PARLAYAPI_KEY":"your_key_here"}}'

Run locally with Docker

Build the stdio container from this repository:

docker build -t parlayapi-mcp:local .
docker run --rm -i --read-only --cap-drop=ALL --security-opt=no-new-privileges parlayapi-mcp:local

The process waits for MCP messages on stdin. It does not open a web server or make an API call just by starting. Connect it through an MCP client; do not use -t, which would allocate a terminal instead of the MCP stdio transport.

For account tools, configure your own PARLAYAPI_KEY in your MCP client's secret configuration and pass that environment variable into the container:

docker run --rm -i --read-only --cap-drop=ALL --security-opt=no-new-privileges -e PARLAYAPI_KEY parlayapi-mcp:local

No key is baked into the image. Each user should configure their own key and keep account responses within their own client. This container is not a shared community feed. The MIT software license grants no data distribution rights; applicable Terms and any written agreement govern API data use. These container defaults do not amend existing agreements.

Public discovery tools need no key. Account tools use your allowance; check current plans and limits. The existing tool set also includes account creation, checkout-link creation, login-link email and saved-book preferences. Approve those actions explicitly in your MCP client. Listing tools itself creates no account, sends no email and starts no checkout.

For private security reports, email support@parlay-api.com. Do not put API keys or account responses in public issues.

Tools exposed

Tool

Auth

Purpose

parlayapi_signup

none

Create a free-tier account, return API key + claim URL + upgrade URL

parlayapi_checkout_link

none

Stripe upgrade URL for any tier

parlayapi_magic_link

none

Email a passwordless login link

parlayapi_get_pricing

none

Current public pricing and credit tiers

parlayapi_live_sports

none

Active live sports, event counts, and book counts

parlayapi_live_search

none

Search live teams, players, and sport keys

parlayapi_live_command_center

none

Public best-line preview from /live

parlayapi_source_quality

none

Public source freshness and quality metadata

parlayapi_book_coverage

none

Public per-book five-gate coverage proof

parlayapi_list_sports

key

All available sport keys

parlayapi_get_odds

key

Game-level odds (h2h, spreads, totals) for a sport

parlayapi_get_props

key

Player prop odds for a sport

parlayapi_best_line

key

Best price per outcome across bookmakers

parlayapi_verdict

key

One-call "should I bet this?": fair price vs market, best book you can bet at, plain-English BET/PASS call

parlayapi_set_bettable_books

key

Remember the user's region/books so verdicts skip books they can't use

parlayapi_parlay_verdict

key

Grade a multi-leg parlay: combined fair price, best book to place it, EV, weakest leg, correlation warnings, payout

parlayapi_best_bets

key

Ranked +EV plays for a sport, scoped to books you can bet at, plus edge alerts

parlayapi_account_info

key

Tier, credits remaining, billing period

parlayapi_find_arbitrage

key

Cross-book arbitrage candidates, 3-way (home/draw/away) markets included

parlayapi_find_ev

key

Positive-EV bets vs a sharp book's no-vig fair line

parlayapi_consensus

key

Consensus (average) odds across all bookmakers per market

parlayapi_find_middles

key

Cross-book middle opportunities

Resources

parlayapi://docs/quickstart - Python and JS code samples.

Local development

git clone https://github.com/JacobiusMakes/parlay-api-mcp
cd parlay-api-mcp
pip install -e .

# Run against the local API instead of production:
PARLAYAPI_BASE_URL=http://127.0.0.1:8080 \
PARLAYAPI_KEY=your_local_test_key \
parlayapi-mcp

Privacy Policy

This extension connects to ParlayAPI. See the ParlayAPI Privacy Policy.

License

MIT


Part of the ParlayAPI ecosystem: a real-time sports odds API with a free tier of 1,000 credits per month, no card required. Explore all the tools at github.com/JacobiusMakes.

Available Tools

22 tools
parlayapi_account_infoAInspect

Get info about the API key currently in use: tier, credits remaining, billing period.

Inputs: no named parameters. Returns: object

parlayapi_best_betsAInspect

What should I bet right now? Ranked +EV plays for a sport.

The discovery counterpart to parlayapi_verdict: scans the whole board, grades every candidate with the same no-vig engine, keeps only bets that are +EV at a book the user can bet at, and ranks them by edge. Also returns edge_alerts (books showing a price far off the market). Player props only, priced by several books, prediction markets and game moneylines excluded (for a game line, use parlayapi_verdict). Scoped to the user's books.

Args: sport_key: e.g. "baseball_mlb". region: us (default) | eu | uk | au | ca. books: exact CSV of books you can bet at (overrides region). limit: max plays to return (default 20). min_edge: minimum edge %% vs the no-vig fair line (default 2.0).

Inputs: sport_key (required), books, limit, region, +1 more. Returns: object

parlayapi_best_lineAInspect

Find the best (highest) price for each outcome across bookmakers.

Useful for line-shopping bots. Returns one row per game / market / side with the bookmaker offering the best price.

Args: sport_key: e.g. "baseball_mlb" market: h2h | spreads | totals bookmakers: Restrict the search to these books (optional).

Inputs: sport_key (required), market, bookmakers. Returns: object

parlayapi_book_coverageBInspect

Return public per-book coverage gates.

No API key required. This is the proof surface for whether each source survives freshness, normalization, database, REST-shape, and stream-shape checks.

Inputs: include_warn, window_minutes. Returns: object

parlayapi_consensusAInspect

Get consensus (average) odds across all bookmakers per market.

Returns average, best, and worst price per (event, market, player, line), a sharp baseline for line-shopping. Soccer and other 3-way markets return separate home, draw, and away consensus rows.

Args: sport_key: e.g. "baseball_mlb", "soccer_epl". markets: Comma-separated market_keys (optional), e.g. "moneyline,totals".

Inputs: sport_key (required), markets. Returns: object

parlayapi_find_arbitrageAInspect

Request calculated cross-book price combinations.

Calculations depend on the returned source quotes and requested filters. Quotes, liquidity and execution can change; profit is not guaranteed. Availability depends on the requested sport, event, market and source.

Args: sport_key: e.g. "baseball_mlb", "soccer_epl". min_profit: Minimum profit % to include (e.g. 1.5 for 1.5%). Default 0. exclude_exchanges: Drop arbs anchored on an exchange (novig/prophetx), whose asks can be no-volume. Default False. markets: Comma-separated market_keys to limit the scan (optional), e.g. "h2h,spreads,totals".

Inputs: sport_key (required), markets, min_profit, exclude_exchanges. Returns: object

parlayapi_find_evAInspect

Find positive-EV bets vs a sharp book's no-vig fair line.

Compares every soft book's price against the no-vig fair probability derived from a sharp book (default Pinnacle); rows where a soft book offers better odds than the sharp fair price are +EV. Three-way soccer markets use a dedicated no-vig pass over home/draw/away, so +EV on the draw surfaces too.

Args: sport_key: e.g. "baseball_mlb", "soccer_epl". sharp_book: Sharp baseline book. Default "pinnacle". min_edge: Minimum edge % to include (e.g. 3 for 3%). Default 2. markets: Comma-separated market_keys (optional).

Inputs: sport_key (required), markets, min_edge, sharp_book. Returns: object

parlayapi_find_middlesAInspect

Find cross-book middle opportunities.

A middle takes the Over at a low line on one book and the Under at a higher line on another, so a window of whole numbers cashes BOTH bets (e.g. Over 7.5 at one book, Under 9.5 at another means 8 or 9 wins both). Scans game totals, spreads, AND player-total props (points, rebounds, strikeouts, ...). Each result carries the window, the numbers that hit, and per-$100 economics: profit_if_hit and net_if_above/below_window.

Args: sport_key: e.g. "baseball_mlb", "basketball_nba". min_gap: Minimum window width in points/runs/goals. Default 1.0. markets: Comma-separated market_keys to limit the scan (optional), e.g. "totals" or "player_points". include_props: Include player-total props alongside game lines. Default True.

Inputs: sport_key (required), markets, min_gap, include_props. Returns: object

parlayapi_get_oddsAInspect

Get game-level odds for a sport from all configured bookmakers.

Args: sport_key: e.g. "baseball_mlb", "soccer_epl" markets: Comma-separated. h2h, spreads, totals. regions: us, us2, uk, eu, fr, au, ca, mx, latam, br, asia. Comma-separated for multiple. bookmakers: Comma-separated bookmaker keys (optional). Examples: pinnacle, draftkings, fanduel, bovada, caesars. odds_format: decimal | american

Inputs: sport_key (required), markets, regions, bookmakers, +1 more. Returns: object

parlayapi_get_pricingAInspect

Return the current public pricing table.

No API key required. Use this when an agent needs to explain which tier unlocks a workflow before sending a checkout link.

Inputs: no named parameters. Returns: object

parlayapi_get_propsAInspect

Get player prop odds for a sport.

Args: sport_key: e.g. "baseball_mlb", "basketball_nba" markets: Comma-separated prop market keys (optional). Examples: player_points, player_rebounds, player_strikeouts, player_passing_yards, player_total_bases. bookmakers: Comma-separated. Examples: prizepicks, underdog, sleeper, draftkings, fanduel. player: Filter to props mentioning this player name. include_futures: Include season-long futures (default False).

Inputs: sport_key (required), player, markets, bookmakers, +1 more. Returns: object

parlayapi_list_sportsAInspect

List available sport keys. Use these in subsequent calls.

Examples: baseball_mlb, basketball_nba, americanfootball_nfl, icehockey_nhl, soccer_epl, mma_ufc, esports_lol, table_tennis_*.

Inputs: active_only. Returns: result

parlayapi_live_command_centerAInspect

Return the public best-line command center payload.

No API key required. Shows current games, participating books, and best available real prices on the /live dashboard preview. This is a discovery and demo surface, not a substitute for the paid odds endpoints.

Inputs: limit, sport_key. Returns: object

parlayapi_live_sportsAInspect

List currently active live sports on the retail /live surface.

No API key required. Returns sport keys plus live event and book counts so agents can discover what is active right now.

Inputs: no named parameters. Returns: result

parlayapi_parlay_verdictAInspect

Grade a multi-leg parlay: should you bet it, and at which book?

Give 2 to 12 legs (each shaped like a parlayapi_verdict bet) and get the combined no-vig fair price, the single best BOOK to place the whole slip at (a real parlay is one slip, not best-of-each), the parlay's EV, the weakest leg, same-game correlation warnings, and (with stake) the payout. Best-price is scoped to books the user can bet at (set once via parlayapi_set_bettable_books, or pass region/books here).

Args: legs: list of leg objects, e.g. [{"sport": "baseball_mlb", "home": "Detroit Tigers", "away": "Kansas City Royals", "market": "h2h", "side": "Detroit Tigers"}, {"sport": "baseball_mlb", "team": "Detroit Tigers", "market": "totals", "side": "over", "line": 8.5}] Each leg: sport, market (h2h/spreads/totals or a player-prop key), side (team, or over/under), home+away or team, player (for props), line. region: us (default) | eu | uk | au | ca. books: exact CSV of books you can bet at (overrides region). stake: optional stake amount, to compute the payout if it hits. book: optional, also report the parlay price at this specific book.

Inputs: legs (required), book, books, kelly, +3 more. Returns: object

parlayapi_set_bettable_booksAInspect

Remember which books the user can bet at, so parlayapi_verdict scopes its best-price and shop recommendations to them automatically (no need to pass region/books every call).

Call this once when a user tells you where they bet ("I'm in the US on DraftKings and FanDuel"). Pass region OR an exact books list.

Args: region: us | eu | uk | au | ca. books: exact CSV of book keys, e.g. "draftkings,fanduel,novig". Overrides region.

Inputs: books, region. Returns: object

parlayapi_signupAInspect

Create a free-tier ParlayAPI account for the given email.

Returns the API key, a magic-login URL the user can click to access their dashboard, and a Stripe upgrade URL. Idempotent: if the email already has an account, returns exists=true with the login URL only (the existing API key is intentionally NOT exposed for security).

Use this when a user wants to start building with ParlayAPI and does not have a key yet. After signup, ask the user to add the returned api_key to their MCP server config under PARLAYAPI_KEY, then restart the MCP client to pick it up.

Args: email: User's email address. intended_use: Optional one-line description of what they're building. Used for analytics. agent_id: Identifier for the agent making the call. Defaults to "mcp-client".

Inputs: email (required), agent_id, intended_use. Returns: object

parlayapi_source_qualityAInspect

Return public per-source freshness and quality metadata.

No API key required. Use this to prove that rows are flowing from each source before choosing books for a customer workflow.

Inputs: limit, minutes. Returns: object

parlayapi_verdictAInspect

Should I bet this? One-call verdict for a specific bet.

Returns the no-vig fair price for the exact bet, the best available price and which book has it, how the user's price (if given) grades as EV, a line-shopping nudge, and a plain-English call: BET / LEAN / FAIR / PASS (or NO_DATA). Use this when a user asks whether a specific bet is worth making, or whether they're getting a good number, instead of stitching /ev + /consensus + /best-line yourself.

The best-price / shop recommendation is scoped to books the user can bet at (default US, so a US bettor is never told to use a euro book). Set it once with parlayapi_set_bettable_books and it is remembered per key, or pass region / books here per call.

Args: sport: sport_key, e.g. "baseball_mlb". side: the team you'd bet (h2h / spreads), or "over" / "under" (totals / props). market: h2h | spreads | totals | a player-prop key (player_hits, player_points, ...). home, away: the two teams, to identify the game. Or pass team with one name. team: one team name to find the game (alternative to home + away). player: player name (required for player-prop markets). line: the number for spreads, totals, or props. book: the book you'd bet at (grades that book's current price). price: the price you're offered (American like -110, or decimal like 1.91). Overrides book, use it to grade a specific number in hand. region: where you can bet: us (default) | eu | uk | au | ca. books: exact CSV of books you can bet at (overrides region; use for state-level geo-blocks, e.g. "draftkings,fanduel,novig"). bankroll: the user's bankroll; when the bet is +EV, returns a suggested Kelly stake amount ("bet it, ~$X"). kelly: Kelly fraction (default 0.5 = half-Kelly, the bankroll-safe standard).

The response also carries an edge_alert when a book shows a price far better than the market (a possible soft error worth grabbing, or a stale line to verify) that the normal best-price guard would otherwise hide.

Inputs: side (required), sport (required), away, book, +10 more. Returns: object

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.8/5.0

    Scored across 22 tools

    Disambiguation4/5

    Most tools have clearly distinct purposes, but the betting analytics cluster (best_line, find_ev, best_bets, verdict, parlay_verdict, consensus) has overlapping goals that could confuse selection despite descriptions. The descriptions do a good job of distinguishing single-bet vs discovery vs parlay, so only minor ambiguity remains.

    Naming Consistency3/5

    All names share a consistent parlayapi_ prefix and snake_case, but the second part mixes verb_noun (get_odds, find_ev, list_sports) with noun phrases (best_line, source_quality, account_info) and bare verbs (signup). This is readable but not a single predictable pattern.

    Tool Count3/5

    22 tools is on the heavy side for the domain, with several analytics tools that overlap in function. The set is comprehensive but could be trimmed; it sits in the borderline 16-25 range.

    Completeness4/5

    The surface covers account lifecycle, odds retrieval, live discovery, and a rich set of betting analytics (EV, arbitrage, middles, verdicts). Minor gaps exist, such as no explicit historical odds or event detail endpoint, but core workflows are fully supported.

    Maintenance

    ActivitySlowing
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      B
      quality
      D
      maintenance
      MCP-compatible server that gives AI agents access to alternative sports data across 30+ leagues — odds, events, probabilities, settlement, and futures for prediction markets, DFS platforms, and sportsbooks.
      29
      29 npm
      37 PyPI
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Live sports betting player props MCP server covering NBA, MLB, NFL, NHL, NCAA, and soccer. Unified from real sportsbooks into one REST API and a real MCP server (Streamable HTTP). Free tier, no card required.
      5
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Props-first sports odds API with a hosted MCP server. Live odds and player props (moneyline, spreads, totals) across US sportsbooks, normalized to JSON. Tools: get_odds, get_props, get_events, get_books. API-key auth, free tier.
      MIT No Attribution
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables retrieval of live multi-sport scores, match video highlights, and match predictions via MCP.
      135 npm
      MIT