NodusAI MCP Server — Prediction Market Signals

NodusAI MCP Server

由 AI 驱动的预测市场信号——任何 AI Agent 都可通过 MCP 访问。

AI Agent 可连接到此服务，获取 Polymarket 和 Kalshi 预测市场的 Oracle 信号。信号由 Gemini 2.5 Flash 结合实时网络信息生成。

---

工作原理

Agent → nodusai.app → connect wallet → pay $1 USDC → get session token
                                                              ↓
Agent → MCP Server (nodus_get_signal) → nodusai.app/api/prediction → signal

1. 访问 nodusai.app 并连接钱包
2. 粘贴 Polymarket 或 Kalshi 市场 URL
3. （可选）添加你期望的结果（YES / NO）
4. 支付 $1 USDC——在链上确认
5. 获取一个可用于 3 次查询的会话令牌
6. 在任何 MCP 客户端中使用会话令牌调用 nodus_get_signal

---

付款模式

• 费用：$1 USDC = 3 次 Oracle 信号查询
• **网络：**Base、Ethereum、Avalanche（任何 EVM 链）
• **代币：**USDC
• **非托管：**付款通过 nodusai.app 直接在链上完成
• **会话：**一次付款 = 一个会话令牌 = 3 次查询（有效期 24 小时）

---

可用工具

工具	描述
nodus_pricing	查看定价以及如何获取会话令牌
nodus_get_signal	使用会话令牌获取 Oracle 信号
nodus_verify_signal	审核过往信号的信息来源
nodus_query_history	你的近期查询历史
nodus_admin_stats	平台整体统计数据（管理员）
nodus_admin_queries	完整查询登记表转储（管理员）

---

信号格式

每条 Oracle 响应都遵循 NodusAI 的结构化架构：

{
  "market_name": "Will the Fed cut rates in June 2026?",
  "predicted_outcome": "YES",
  "probability": 0.73,
  "confidence_score": "HIGH",
  "key_reasoning": "Recent FOMC minutes and inflation data suggest...",
  "grounding_sources": [
    { "title": "Reuters: Fed signals rate path", "url": "https://..." },
    { "title": "AP: CPI data June 2026", "url": "https://..." }
  ]
}

---

5 分钟内完成部署

选项 1 — Railway（推荐）

1. 在 GitHub 上 Fork 此仓库
2. 前往 railway.app → New Project → Deploy from GitHub repo
3. 选择你的 Fork
4. 添加环境变量：NODUSAI_API_BASE = https://nodusai.app
5. Railway 会自动检测 railway.json 并进行部署
6. 复制你的 Railway URL

---

选项 2 — Render（免费套餐）

1. Fork 此仓库
2. 前往 render.com → New Web Service → 连接你的 Fork
3. 设置构建命令：npm install 和启动命令：node src/server-http.js
4. 添加环境变量：NODUSAI_API_BASE=https://nodusai.app

---

选项 3 — Fly.io

fly launch --name nodusai-mcp
fly secrets set NODUSAI_API_BASE=https://nodusai.app
fly deploy

---

连接 AI Agent

Claude Desktop

文件：~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "nodusai": {
      "url": "https://nodusai-mcp-production.up.railway.app/sse"
    }
  }
}

---

Cursor

文件：~/.cursor/mcp.json

{
  "mcpServers": {
    "nodusai": {
      "url": "https://nodusai-mcp-production.up.railway.app/sse",
      "transport": "sse"
    }
  }
}

---

Windsurf

文件：~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "nodusai": {
      "serverUrl": "https://nodusai-mcp-production.up.railway.app/sse"
    }
  }
}

---

Claude Code（CLI）

claude mcp add --transport sse nodusai https://nodusai-mcp-production.up.railway.app/sse

---

自定义 JS Agent

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { SSEClientTransport } from "@modelcontextprotocol/sdk/client/sse.js";

const client = new Client({ name: "my-agent", version: "1.0.0" }, { capabilities: {} });
await client.connect(new SSEClientTransport(new URL("https://nodusai-mcp-production.up.railway.app/sse")));

// Step 1 — get a session token at https://nodusai.app ($1 USDC)

// Step 2 — query the Oracle
const result = await client.callTool({
  name: "nodus_get_signal",
  arguments: {
    marketUrl:      "https://polymarket.com/event/...",
    sessionToken:   "your-session-token-from-nodusai.app",
    desiredOutcome: "YES", // optional
  }
});

---

自定义 Python Agent

from mcp.client.sse import sse_client
from mcp import ClientSession

async with sse_client("https://nodusai-mcp-production.up.railway.app/sse") as (read, write):
    async with ClientSession(read, write) as session:
        await session.initialize()

        # Get a session token at https://nodusai.app first ($1 USDC)
        result = await session.call_tool("nodus_get_signal", {
            "marketUrl":      "https://kalshi.com/markets/...",
            "sessionToken":   "your-session-token-from-nodusai.app",
            "desiredOutcome": "YES",  # optional
        })

---

本地开发

git clone https://github.com/NodusAI-Your-Prediction-Broker/nodusai-mcp
cd nodusai-mcp
npm install

# Dev mode (mock oracle — no real API calls needed)
npm run dev:http

测试：

curl http://localhost:3000/health
curl http://localhost:3000/info