Skip to main content
Glama
ledgerfc-mcp

ledgerfc-mcp

Ledger FC MCP

为你的 AI 智能体提供真实的体育数据,而不是不可靠的猜测。

当你向 AI 询问一场比赛,它往往只会依据过时的训练数据作答。这个 MCP 提供的则是实时、有据可依的数字:每个赛果对应的模型概率、它们与市场赔率的分歧点,以及这些预测最终表现如何的公开追踪记录——胜负均在册。另有欧洲足球和网球的实时套利,覆盖了那些美国专属的体育 MCP 未触及的空白。

以诚实为本

这不是一款“保证盈利”的工具。每一项响应都附带一个 provenance 块和一条 for_the_agent 说明(“可供参考的基础数据,而非行动指令”)。实时拉取的真实数字如下:

  • 整体模型准确率:约 53%(几乎和扔硬币差不多)。我们会如实相告。

  • 高置信度子集:约 68%(96 个已结算结果中命中 65 个)。

  • 收盘线价值(CLV):基本持平。我们会返回该指标,并加以标记,因此你的智能体只有在样本足够大时才会把它纳入权衡。

如果你想要的是一款基于事实而非夸夸其谈的智能体,恰恰正是这种诚实。

Related MCP server: SportsQuant MCP Server

覆盖范围

欧洲足球(EPL、La Liga、Serie A、Ligue 1、Eredivisie、Turkish Super Lig)、网球(ATP),外加 MLB、NFL、NBA。目前主流的免费体育博彩 MCP 仅限美国,所以欧洲足球的套利正是本工具所填补的空白。

安装

Claude Desktop / Claude Code / Cursor(通过 npx 使用 stdio)

{
  "mcpServers": {
    "ledgerfc": {
      "command": "npx",
      "args": ["-y", "@inbin/ledgerfc-mcp"]
    }
  }
}

免费工具无需 API key,就这样即可。

如需同时使用 Pro 工具,则将 key 设为环境变量即可(在 Ledger FC Telegram 机器人中用 /key 获取)。该 key 通过 Authorization 请求头发送,不会进入对话内容:

{
  "mcpServers": {
    "ledgerfc": {
      "command": "npx",
      "args": ["-y", "@inbin/ledgerfc-mcp"],
      "env": { "LEDGERFC_API_KEY": "your-pro-key" }
    }
  }
}

也可直接连接远程服务(支持 remote MCP 的客户端)

https://mcp.ledgerfc.com/mcp

对于 Pro 工具,请将 key 作为 Authorization: Bearer <key> 请求头发送。

工具

免费(无需 key):

  • get_track_record:真实的已结算记录,另有高置信度子集与 CLV

  • get_predictions:即将到来的赛果预测,含概率及模型对市场的空间差(可按联赛筛选)

  • search / fetch:查询即将进行的比赛预测,拉取完整文档

  • get_leaderboard、submit_prediction、get_contributor_score:按 CLV 计分的贡献者层(计算的是积分,不是真钱)

  • place_bet、get_balance、get_my_bets、get_bet:模拟下注账本,用于对我们是智能体测试策略(仅有积分)

Pro(设置 LEDGERFC_API_KEY,或将 api_key 作为工具参数传入;可在 Ledger FC Telegram bot 中用 /key 获取):

  • get_arbs:实时探测到的套利机会

  • get_value_bets:即将到来且模型对市场具有正优势的比赛

示例

向智能体提问:“用 ledgerfc 获取历史上的战绩和优势最大的前 3 场即将到来的比赛预测。” 它会调用 get_track_record 和 get_predictions,然后给你一个有真凭实据的答案,而不是瞎猜。

工作原理

这个 npm 包其实是一个轻量的 stdio-to-HTTP 桥接器:它从 stdin 读取你 MCP 客户端传入的 JSON-RPC 请求,并将逐个请求转发到远程端点 https://mcp.ledgerfc.com/mcp。零依赖,Node 18+ 即可使用。设 LEDGERFC_MCP_DEBUG=1 能把流量日志打到 stderr,或 LEDGERFC_MCP_URL 可指定不同的端点。

相关链接

MIT 开源协议。Ledger FC 是信息工具,不构成下注建议。18+。

Available Tools

13 tools
fetchBInspect

Fetch the full prediction document for a search result id.

Inputs: id (required).

get_arbsCInspect

PRO: recently detected live arbitrage opportunities. Needs a Pro key, via an Authorization: Bearer header (preferred) or the api_key argument.

Inputs: api_key, live_rescan, within_seconds.

get_balanceCInspect

Your paper-points balance (not cash).

Inputs: agent_id (required).

get_betCInspect

One paper bet by id.

Inputs: bet_id (required).

get_contributor_scoreDInspect

A contributor's CLV record.

Inputs: contributor_id (required).

get_leaderboardBInspect

Contributors ranked by CLV (reputation points, not cash).

Inputs: no named parameters.

get_my_betsCInspect

Your paper bets, newest first.

Inputs: agent_id (required).

get_predictionsBInspect

Upcoming predictions with probabilities and model-vs-market edge.

Inputs: limit, league.

get_track_recordDInspect

Honest settled record + high-confidence subset + CLV.

Inputs: no named parameters.

get_value_betsAInspect

PRO: upcoming picks with positive model-vs-market edge. Needs a Pro key, via an Authorization: Bearer header (preferred) or the api_key argument.

Inputs: api_key.

place_betCInspect

Place a paper bet against our agent's prediction (points only).

Inputs: side (required), agent_id (required), match_id (required), stake_points (required).

submit_predictionCInspect

Submit a prediction to be scored by CLV (shadow/points).

Inputs: p_away (required), p_draw (required), p_home (required), match_id (required), +1 more.

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    B3/5.0

    Scored across 13 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose: track record, predictions, arbs, value bets, search/fetch pair, submission, leaderboard, contributor score, betting operations. No two tools overlap in a way that would cause misselection.

    Naming Consistency5/5

    All tools use consistent lowercase snake_case with a verb_noun pattern. Most start with 'get_', with a few action verbs like 'search', 'fetch', 'submit_prediction', and 'place_bet' that still follow the same style.

    Tool Count5/5

    13 tools is well within the ideal range for a domain covering predictions, submission, scoring, and betting. Each tool serves a distinct function and none appear redundant.

    Completeness4/5

    The surface covers the main lifecycle: viewing predictions, fetching details, submitting predictions, checking leaderboard and contributor scores, placing bets, and viewing balance and bet history. Minor missing operations like updating or canceling bets are not critical for the core workflow.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      B
      maintenance
      Live and pre-match sports odds from 450+ bookmakers across 34 sports, plus Polymarket and Kalshi. Events, odds, fair prices, value bets, arbitrage, historical closing lines and bet settlement for AI assistants.
      33
      1,129 npm
      2
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables AI agents to query live sports betting data including odds, edges, arbitrage opportunities, player/team stats, and reference data using the SportWizzard API.
      20
      132 npm
      MIT
    • A
      license
      A
      quality
      C
      maintenance
      Provides AI assistants with sports model win probabilities and fair odds across nine sports without requiring an API key.
      3
      54 npm
      MIT