diff --git a/backend/app/config.py b/backend/app/config.py
index ddf291e..12bdb57 100644
--- a/backend/app/config.py
+++ b/backend/app/config.py
@@ -89,6 +89,21 @@ class Settings(BaseSettings):
     # CORS配置
     cors_origins: str = "http://localhost:8000,http://127.0.0.1:8000"
 
+    # Binance 配置（公开数据不需要 API 密钥）
+    binance_api_key: str = ""
+    binance_api_secret: str = ""
+
+    # 飞书机器人配置
+    feishu_webhook_url: str = "https://open.feishu.cn/open-apis/bot/v2/hook/8a1dcf69-6753-41e2-a393-edc4f7822db0"
+
+    # 加密货币交易智能体配置
+    crypto_symbols: str = "BTCUSDT,ETHUSDT"  # 监控的交易对，逗号分隔
+    crypto_analysis_interval: int = 60  # 分析间隔（秒）
+    crypto_llm_threshold: float = 0.7  # 触发 LLM 分析的置信度阈值
+
+    # Brave Search API 配置
+    brave_api_key: str = ""
+
     class Config:
         env_file = find_env_file()
         case_sensitive = False
diff --git a/backend/app/crypto_agent/__init__.py b/backend/app/crypto_agent/__init__.py
new file mode 100644
index 0000000..30c8a09
--- /dev/null
+++ b/backend/app/crypto_agent/__init__.py
@@ -0,0 +1,8 @@
+"""
+加密货币交易智能体模块
+"""
+from app.crypto_agent.crypto_agent import CryptoAgent
+from app.crypto_agent.signal_analyzer import SignalAnalyzer
+from app.crypto_agent.strategy import TrendFollowingStrategy
+
+__all__ = ['CryptoAgent', 'SignalAnalyzer', 'TrendFollowingStrategy']
diff --git a/backend/app/crypto_agent/crypto_agent.py b/backend/app/crypto_agent/crypto_agent.py
new file mode 100644
index 0000000..c5d3e57
--- /dev/null
+++ b/backend/app/crypto_agent/crypto_agent.py
@@ -0,0 +1,258 @@
+"""
+加密货币交易智能体 - 主控制器
+"""
+import asyncio
+from typing import Dict, Any, List, Optional
+from datetime import datetime, timedelta
+import pandas as pd
+
+from app.utils.logger import logger
+from app.config import get_settings
+from app.services.binance_service import binance_service
+from app.services.feishu_service import get_feishu_service
+from app.crypto_agent.signal_analyzer import SignalAnalyzer
+from app.crypto_agent.strategy import TrendFollowingStrategy
+
+
+class CryptoAgent:
+    """加密货币交易信号智能体"""
+
+    def __init__(self):
+        """初始化智能体"""
+        self.settings = get_settings()
+        self.binance = binance_service
+        self.feishu = get_feishu_service()
+        self.analyzer = SignalAnalyzer()
+        self.strategy = TrendFollowingStrategy()
+
+        # 状态管理
+        self.last_signals: Dict[str, Dict[str, Any]] = {}  # 上次信号
+        self.last_trends: Dict[str, str] = {}  # 上次趋势
+        self.signal_cooldown: Dict[str, datetime] = {}  # 信号冷却时间
+
+        # 配置
+        self.symbols = self.settings.crypto_symbols.split(',')
+        self.analysis_interval = self.settings.crypto_analysis_interval
+        self.llm_threshold = self.settings.crypto_llm_threshold
+
+        # 运行状态
+        self.running = False
+
+        logger.info(f"加密货币智能体初始化完成，监控交易对: {self.symbols}")
+
+    async def run(self):
+        """主运行循环"""
+        self.running = True
+        logger.info("加密货币智能体开始运行...")
+
+        # 发送启动通知
+        await self.feishu.send_text(
+            f"🚀 加密货币智能体已启动\n"
+            f"监控交易对: {', '.join(self.symbols)}\n"
+            f"分析间隔: {self.analysis_interval}秒"
+        )
+
+        while self.running:
+            try:
+                for symbol in self.symbols:
+                    await self.analyze_symbol(symbol)
+
+                # 等待下一次分析
+                await asyncio.sleep(self.analysis_interval)
+
+            except Exception as e:
+                logger.error(f"分析循环出错: {e}")
+                await asyncio.sleep(10)  # 出错后等待10秒再继续
+
+    def stop(self):
+        """停止运行"""
+        self.running = False
+        logger.info("加密货币智能体已停止")
+
+    async def analyze_symbol(self, symbol: str):
+        """
+        分析单个交易对
+
+        Args:
+            symbol: 交易对，如 'BTCUSDT'
+        """
+        try:
+            logger.info(f"开始分析 {symbol}...")
+
+            # 1. 获取多周期数据
+            data = self.binance.get_multi_timeframe_data(symbol)
+
+            if not self._validate_data(data):
+                logger.warning(f"{symbol} 数据不完整，跳过分析")
+                return
+
+            # 2. 分析趋势（1H + 4H）- 返回详细趋势信息
+            trend = self.analyzer.analyze_trend(data['1h'], data['4h'])
+            trend_direction = trend.get('direction', 'neutral') if isinstance(trend, dict) else trend
+
+            # 3. 检查趋势变化
+            last_direction = self.last_trends.get(symbol, {})
+            if isinstance(last_direction, dict):
+                last_direction = last_direction.get('direction', 'neutral')
+            if last_direction and last_direction != trend_direction:
+                await self._handle_trend_change(symbol, last_direction, trend_direction, data)
+
+            self.last_trends[symbol] = trend
+
+            # 4. 分析进场信号（15M 为主，5M 辅助）
+            signal = self.analyzer.analyze_entry_signal(data['5m'], data['15m'], trend)
+            signal['symbol'] = symbol
+            signal['trend'] = trend_direction
+            signal['trend_info'] = trend if isinstance(trend, dict) else {'direction': trend}
+            signal['price'] = float(data['5m'].iloc[-1]['close'])
+            signal['timestamp'] = datetime.now()
+
+            # 5. 检查是否需要发送信号
+            if self._should_send_signal(symbol, signal):
+                # 6. 计算止损止盈
+                atr = float(data['15m'].iloc[-1].get('atr', 0))
+                if atr > 0:
+                    sl_tp = self.analyzer.calculate_stop_loss_take_profit(
+                        signal['price'], signal['action'], atr
+                    )
+                    signal.update(sl_tp)
+
+                # 7. LLM 深度分析（置信度超过阈值时）
+                if signal['confidence'] >= self.llm_threshold * 100:
+                    llm_result = await self.analyzer.llm_analyze(data, signal, symbol)
+
+                    # 处理 LLM 分析结果
+                    if llm_result.get('parsed'):
+                        parsed = llm_result['parsed']
+                        # 新格式使用 signal 而不是 recommendation
+                        recommendation = parsed.get('signal', parsed.get('recommendation', {}))
+
+                        # 如果 LLM 建议观望，降低置信度
+                        if recommendation.get('action') == 'wait':
+                            signal['confidence'] = min(signal['confidence'], 40)
+                            signal['llm_analysis'] = llm_result.get('summary', 'LLM 建议观望')
+                        else:
+                            # 使用 LLM 的止损止盈建议
+                            if recommendation.get('stop_loss'):
+                                signal['stop_loss'] = recommendation['stop_loss']
+                            if recommendation.get('targets'):
+                                signal['take_profit'] = recommendation['targets'][0]
+                            elif recommendation.get('take_profit'):
+                                signal['take_profit'] = recommendation['take_profit']
+                            signal['llm_analysis'] = llm_result.get('summary', '')
+                    else:
+                        signal['llm_analysis'] = llm_result.get('summary', llm_result.get('raw', '')[:200])
+
+                # 8. 发送飞书通知（置信度仍然足够高时）
+                if signal['confidence'] >= 50:
+                    await self.feishu.send_trading_signal(signal)
+
+                    # 9. 更新状态
+                    self.last_signals[symbol] = signal
+                    self.signal_cooldown[symbol] = datetime.now()
+
+                    logger.info(f"{symbol} 发送{signal['action']}信号，置信度: {signal['confidence']}%")
+
+        except Exception as e:
+            logger.error(f"分析 {symbol} 出错: {e}")
+
+    def _validate_data(self, data: Dict[str, pd.DataFrame]) -> bool:
+        """验证数据完整性"""
+        required_intervals = ['5m', '15m', '1h', '4h']
+        for interval in required_intervals:
+            if interval not in data or data[interval].empty:
+                return False
+            if len(data[interval]) < 20:  # 至少需要20条数据
+                return False
+        return True
+
+    def _should_send_signal(self, symbol: str, signal: Dict[str, Any]) -> bool:
+        """
+        判断是否应该发送信号
+
+        Args:
+            symbol: 交易对
+            signal: 信号数据
+
+        Returns:
+            是否发送
+        """
+        # 如果是观望，不发送
+        if signal['action'] == 'hold':
+            return False
+
+        # 置信度太低，不发送
+        if signal['confidence'] < 50:
+            return False
+
+        # 检查冷却时间（同一交易对30分钟内不重复发送相同方向的信号）
+        if symbol in self.signal_cooldown:
+            cooldown_end = self.signal_cooldown[symbol] + timedelta(minutes=30)
+            if datetime.now() < cooldown_end:
+                # 检查是否是相同方向的信号
+                if symbol in self.last_signals:
+                    if self.last_signals[symbol]['action'] == signal['action']:
+                        logger.debug(f"{symbol} 信号冷却中，跳过")
+                        return False
+
+        return True
+
+    async def _handle_trend_change(self, symbol: str, old_trend: str, new_trend: str,
+                                   data: Dict[str, pd.DataFrame]):
+        """处理趋势变化"""
+        price = float(data['1h'].iloc[-1]['close'])
+        await self.feishu.send_trend_change(symbol, old_trend, new_trend, price)
+        logger.info(f"{symbol} 趋势变化: {old_trend} -> {new_trend}")
+
+    async def analyze_once(self, symbol: str) -> Dict[str, Any]:
+        """
+        单次分析（用于测试或手动触发）
+
+        Args:
+            symbol: 交易对
+
+        Returns:
+            分析结果
+        """
+        data = self.binance.get_multi_timeframe_data(symbol)
+
+        if not self._validate_data(data):
+            return {'error': '数据不完整'}
+
+        trend = self.analyzer.analyze_trend(data['1h'], data['4h'])
+        signal = self.analyzer.analyze_entry_signal(data['5m'], data['15m'], trend)
+
+        signal['symbol'] = symbol
+        signal['trend'] = trend
+        signal['price'] = float(data['5m'].iloc[-1]['close'])
+
+        # 计算止损止盈
+        atr = float(data['15m'].iloc[-1].get('atr', 0))
+        if atr > 0:
+            sl_tp = self.analyzer.calculate_stop_loss_take_profit(
+                signal['price'], signal['action'], atr
+            )
+            signal.update(sl_tp)
+
+        return signal
+
+    def get_status(self) -> Dict[str, Any]:
+        """获取智能体状态"""
+        return {
+            'running': self.running,
+            'symbols': self.symbols,
+            'analysis_interval': self.analysis_interval,
+            'last_signals': {
+                symbol: {
+                    'action': sig.get('action'),
+                    'confidence': sig.get('confidence'),
+                    'timestamp': sig.get('timestamp').isoformat() if sig.get('timestamp') else None
+                }
+                for symbol, sig in self.last_signals.items()
+            },
+            'last_trends': self.last_trends
+        }
+
+
+# 全局实例
+crypto_agent = CryptoAgent()
diff --git a/backend/app/crypto_agent/signal_analyzer.py b/backend/app/crypto_agent/signal_analyzer.py
new file mode 100644
index 0000000..d3ec3c7
--- /dev/null
+++ b/backend/app/crypto_agent/signal_analyzer.py
@@ -0,0 +1,670 @@
+"""
+信号分析器 - 多周期技术分析和 LLM 深度分析
+"""
+import pandas as pd
+from typing import Dict, Any, Optional, List
+from app.utils.logger import logger
+from app.services.llm_service import llm_service
+
+
+class SignalAnalyzer:
+    """交易信号分析器 - 波段交易优化版"""
+
+    # LLM 系统提示词 - 波段交易版
+    CRYPTO_ANALYST_PROMPT = """你是一位经验丰富的加密货币波段交易员，专注于捕捉 1-7 天的中等波段行情。
+
+## 交易风格
+- **波段交易**：持仓 1-7 天，不做超短线
+- **顺势回调**：在趋势中寻找回调入场机会
+- **风险控制**：单笔亏损不超过本金 2%
+
+## 多周期分析框架
+1. **4H 周期**：判断主趋势方向和强度
+   - 趋势明确：价格在 MA20 同侧运行 3 根以上 K 线
+   - 趋势强度：看 MACD 柱状图是否放大
+
+2. **1H 周期**：确认趋势 + 寻找回调位置
+   - 上涨趋势中：等待回调到 MA20 或前低支撑
+   - 下跌趋势中：等待反弹到 MA20 或前高阻力
+
+3. **15M 周期**：入场信号确认
+   - 做多：RSI 从超卖回升 + MACD 金叉 + K 线企稳
+   - 做空：RSI 从超买回落 + MACD 死叉 + K 线见顶
+
+## 入场条件（波段做多）
+1. 4H 趋势向上（价格 > MA20，MACD > 0 或底背离）
+2. 1H 回调到支撑位（MA20 附近或前低）
+3. 15M 出现止跌信号（RSI < 40 回升，或 MACD 金叉）
+4. 止损明确（前低下方），风险收益比 >= 1:2
+
+## 入场条件（波段做空）
+1. 4H 趋势向下（价格 < MA20，MACD < 0 或顶背离）
+2. 1H 反弹到阻力位（MA20 附近或前高）
+3. 15M 出现见顶信号（RSI > 60 回落，或 MACD 死叉）
+4. 止损明确（前高上方），风险收益比 >= 1:2
+
+## 特殊情况处理
+- **极度超卖（RSI < 20）**：不追空，等待反弹做多机会
+- **极度超买（RSI > 80）**：不追多，等待回调做空机会
+- **震荡市**：观望，等待突破方向
+
+## 输出格式（JSON）
+```json
+{
+  "market_structure": {
+    "trend": "uptrend/downtrend/sideways",
+    "strength": "strong/moderate/weak",
+    "phase": "impulse/correction/reversal"
+  },
+  "key_levels": {
+    "resistance": [阻力位1, 阻力位2],
+    "support": [支撑位1, 支撑位2]
+  },
+  "signal": {
+    "quality": "A/B/C/D",
+    "action": "buy/sell/wait",
+    "confidence": 0-100,
+    "entry_zone": [入场区间下限, 入场区间上限],
+    "stop_loss": 止损价,
+    "targets": [目标1, 目标2],
+    "reason": "入场理由"
+  },
+  "risk_warning": "风险提示"
+}
+```
+
+信号质量说明：
+- A级：趋势明确 + 回调到位 + 多重信号共振（置信度 80+）
+- B级：趋势明确 + 信号较好（置信度 60-80）
+- C级：有机会但需要更多确认（置信度 40-60）
+- D级：不建议交易（置信度 < 40）
+
+重要：波段交易要有耐心，宁可错过也不要在不理想的位置入场。"""
+
+    def __init__(self):
+        """初始化信号分析器"""
+        logger.info("信号分析器初始化完成")
+
+    def analyze_trend(self, h1_data: pd.DataFrame, h4_data: pd.DataFrame) -> Dict[str, Any]:
+        """
+        分析趋势方向和强度（波段交易优化版）
+
+        Args:
+            h1_data: 1小时K线数据（含技术指标）
+            h4_data: 4小时K线数据（含技术指标）
+
+        Returns:
+            {
+                'direction': 'bullish' | 'bearish' | 'neutral',
+                'strength': 'strong' | 'moderate' | 'weak',
+                'phase': 'impulse' | 'correction' | 'reversal',
+                'h4_score': float,
+                'h1_score': float
+            }
+        """
+        if h1_data.empty or h4_data.empty:
+            return {
+                'direction': 'neutral',
+                'strength': 'weak',
+                'phase': 'sideways',
+                'h4_score': 0,
+                'h1_score': 0
+            }
+
+        # 获取最新数据
+        h1_latest = h1_data.iloc[-1]
+        h4_latest = h4_data.iloc[-1]
+
+        # 计算各周期的趋势得分
+        h4_score, h4_details = self._calculate_trend_score(h4_latest)
+        h1_score, h1_details = self._calculate_trend_score(h1_latest)
+
+        # 判断趋势方向（4H 为主）
+        if h4_score > 0.3:
+            direction = 'bullish'
+        elif h4_score < -0.3:
+            direction = 'bearish'
+        else:
+            direction = 'neutral'
+
+        # 判断趋势强度
+        strength = self._assess_trend_strength(h4_data, h1_data)
+
+        # 判断当前阶段（是主升/主跌还是回调）
+        phase = self._detect_market_phase(h4_data, h1_data, direction)
+
+        # 检查极端情况
+        h4_rsi = h4_latest.get('rsi', 50)
+        extreme_warning = ""
+        if pd.notna(h4_rsi):
+            if h4_rsi < 20:
+                extreme_warning = f" [RSI={h4_rsi:.1f} 极度超卖]"
+                phase = 'oversold'
+            elif h4_rsi > 80:
+                extreme_warning = f" [RSI={h4_rsi:.1f} 极度超买]"
+                phase = 'overbought'
+
+        logger.info(f"趋势分析: 方向={direction}, 强度={strength}, 阶段={phase} | "
+                    f"4H={h4_score:.2f}{h4_details}, 1H={h1_score:.2f}{h1_details}{extreme_warning}")
+
+        return {
+            'direction': direction,
+            'strength': strength,
+            'phase': phase,
+            'h4_score': h4_score,
+            'h1_score': h1_score
+        }
+
+    def _assess_trend_strength(self, h4_data: pd.DataFrame, h1_data: pd.DataFrame) -> str:
+        """评估趋势强度"""
+        if len(h4_data) < 5:
+            return 'weak'
+
+        h4_latest = h4_data.iloc[-1]
+
+        # 检查 MACD 柱状图是否放大
+        macd_hist = h4_data['macd_hist'].tail(5)
+        macd_expanding = False
+        if len(macd_hist) >= 3:
+            recent_abs = abs(macd_hist.iloc[-1])
+            prev_abs = abs(macd_hist.iloc[-3])
+            if recent_abs > prev_abs * 1.2:
+                macd_expanding = True
+
+        # 检查价格是否持续在 MA20 同侧
+        close_prices = h4_data['close'].tail(5)
+        ma20_values = h4_data['ma20'].tail(5)
+        consistent_side = True
+        if pd.notna(ma20_values.iloc[-1]):
+            above_ma = close_prices > ma20_values
+            consistent_side = above_ma.all() or (~above_ma).all()
+
+        # 检查 RSI 是否在趋势区间
+        rsi = h4_latest.get('rsi', 50)
+        rsi_trending = 40 < rsi < 60  # 中性区间表示趋势不强
+
+        if macd_expanding and consistent_side and not rsi_trending:
+            return 'strong'
+        elif consistent_side:
+            return 'moderate'
+        else:
+            return 'weak'
+
+    def _detect_market_phase(self, h4_data: pd.DataFrame, h1_data: pd.DataFrame,
+                             direction: str) -> str:
+        """检测市场阶段（主升/主跌 vs 回调）"""
+        if len(h1_data) < 10:
+            return 'unknown'
+
+        h1_latest = h1_data.iloc[-1]
+        h4_latest = h4_data.iloc[-1]
+
+        # 获取 1H 的短期趋势
+        h1_ma5 = h1_latest.get('ma5', 0)
+        h1_ma20 = h1_latest.get('ma20', 0)
+        h1_close = h1_latest.get('close', 0)
+
+        if not (pd.notna(h1_ma5) and pd.notna(h1_ma20)):
+            return 'unknown'
+
+        # 判断 1H 是否在回调
+        if direction == 'bullish':
+            # 上涨趋势中，1H 价格回落到 MA20 附近 = 回调
+            if h1_close < h1_ma5 and h1_close > h1_ma20 * 0.98:
+                return 'correction'  # 回调中，可能是入场机会
+            elif h1_close > h1_ma5:
+                return 'impulse'  # 主升浪
+        elif direction == 'bearish':
+            # 下跌趋势中，1H 价格反弹到 MA20 附近 = 反弹
+            if h1_close > h1_ma5 and h1_close < h1_ma20 * 1.02:
+                return 'correction'  # 反弹中，可能是做空机会
+            elif h1_close < h1_ma5:
+                return 'impulse'  # 主跌浪
+
+        return 'sideways'
+
+    def _calculate_trend_score(self, data: pd.Series) -> tuple:
+        """
+        计算单周期趋势得分
+
+        Args:
+            data: 包含技术指标的数据行
+
+        Returns:
+            (得分, 详情字符串)
+        """
+        score = 0.0
+        count = 0
+        details = []
+
+        # 价格与均线关系
+        if 'close' in data and 'ma20' in data and pd.notna(data['ma20']):
+            if data['close'] > data['ma20']:
+                score += 1
+                details.append("价格>MA20")
+            else:
+                score -= 1
+                details.append("价格<MA20")
+            count += 1
+
+        # MA5 与 MA20 关系
+        if 'ma5' in data and 'ma20' in data and pd.notna(data['ma5']) and pd.notna(data['ma20']):
+            if data['ma5'] > data['ma20']:
+                score += 1
+                details.append("MA5>MA20")
+            else:
+                score -= 1
+                details.append("MA5<MA20")
+            count += 1
+
+        # MACD
+        if 'macd' in data and 'macd_signal' in data and pd.notna(data['macd']):
+            if data['macd'] > data['macd_signal']:
+                score += 1
+                details.append("MACD多")
+            else:
+                score -= 1
+                details.append("MACD空")
+            count += 1
+
+        # RSI
+        if 'rsi' in data and pd.notna(data['rsi']):
+            if data['rsi'] > 50:
+                score += 0.5
+            else:
+                score -= 0.5
+            count += 0.5
+
+        final_score = score / count if count > 0 else 0
+        detail_str = f"({','.join(details)})" if details else ""
+        return final_score, detail_str
+
+    def analyze_entry_signal(self, m5_data: pd.DataFrame, m15_data: pd.DataFrame,
+                             trend: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        分析 15M 进场信号（波段交易优化版）
+
+        Args:
+            m5_data: 5分钟K线数据（用于精确入场）
+            m15_data: 15分钟K线数据（主要入场周期）
+            trend: 趋势分析结果
+
+        Returns:
+            {
+                'action': 'buy' | 'sell' | 'hold',
+                'confidence': 0-100,
+                'signal_grade': 'A' | 'B' | 'C' | 'D',
+                'reasons': [...],
+                'indicators': {...}
+            }
+        """
+        if m5_data.empty or m15_data.empty:
+            return {'action': 'hold', 'confidence': 0, 'signal_grade': 'D',
+                    'reasons': ['数据不足'], 'indicators': {}}
+
+        # 兼容旧格式（如果 trend 是字符串）
+        if isinstance(trend, str):
+            trend_direction = trend
+            trend_phase = 'unknown'
+            trend_strength = 'moderate'
+        else:
+            trend_direction = trend.get('direction', 'neutral')
+            trend_phase = trend.get('phase', 'unknown')
+            trend_strength = trend.get('strength', 'moderate')
+
+        m15_latest = m15_data.iloc[-1]
+
+        # 收集信号
+        buy_signals = []
+        sell_signals = []
+        signal_weights = {'buy': 0, 'sell': 0}
+
+        # === RSI 信号 ===
+        if 'rsi' in m15_latest and pd.notna(m15_latest['rsi']):
+            rsi = m15_latest['rsi']
+            if rsi < 30:
+                buy_signals.append(f"RSI超卖({rsi:.1f})")
+                signal_weights['buy'] += 2
+            elif rsi < 40 and len(m15_data) >= 2:
+                # RSI 从低位回升
+                prev_rsi = m15_data.iloc[-2].get('rsi', 50)
+                if pd.notna(prev_rsi) and rsi > prev_rsi:
+                    buy_signals.append(f"RSI回升({prev_rsi:.1f}→{rsi:.1f})")
+                    signal_weights['buy'] += 1.5
+            elif rsi > 70:
+                sell_signals.append(f"RSI超买({rsi:.1f})")
+                signal_weights['sell'] += 2
+            elif rsi > 60 and len(m15_data) >= 2:
+                prev_rsi = m15_data.iloc[-2].get('rsi', 50)
+                if pd.notna(prev_rsi) and rsi < prev_rsi:
+                    sell_signals.append(f"RSI回落({prev_rsi:.1f}→{rsi:.1f})")
+                    signal_weights['sell'] += 1.5
+
+        # === MACD 信号 ===
+        if len(m15_data) >= 2:
+            prev = m15_data.iloc[-2]
+            if 'macd' in m15_latest and 'macd_signal' in m15_latest:
+                if pd.notna(m15_latest['macd']) and pd.notna(prev['macd']):
+                    # 金叉
+                    if prev['macd'] <= prev['macd_signal'] and m15_latest['macd'] > m15_latest['macd_signal']:
+                        buy_signals.append("MACD金叉")
+                        signal_weights['buy'] += 2
+                    # 死叉
+                    elif prev['macd'] >= prev['macd_signal'] and m15_latest['macd'] < m15_latest['macd_signal']:
+                        sell_signals.append("MACD死叉")
+                        signal_weights['sell'] += 2
+                    # MACD 柱状图缩小（趋势减弱）
+                    elif abs(m15_latest['macd_hist']) < abs(prev['macd_hist']) * 0.7:
+                        if m15_latest['macd_hist'] > 0:
+                            sell_signals.append("MACD动能减弱")
+                            signal_weights['sell'] += 0.5
+                        else:
+                            buy_signals.append("MACD动能减弱")
+                            signal_weights['buy'] += 0.5
+
+        # === 布林带信号 ===
+        if 'close' in m15_latest and 'bb_lower' in m15_latest and 'bb_upper' in m15_latest:
+            if pd.notna(m15_latest['bb_lower']) and pd.notna(m15_latest['bb_upper']):
+                bb_middle = m15_latest.get('bb_middle', (m15_latest['bb_upper'] + m15_latest['bb_lower']) / 2)
+                if m15_latest['close'] < m15_latest['bb_lower']:
+                    buy_signals.append("触及布林下轨")
+                    signal_weights['buy'] += 1.5
+                elif m15_latest['close'] > m15_latest['bb_upper']:
+                    sell_signals.append("触及布林上轨")
+                    signal_weights['sell'] += 1.5
+                # 突破中轨
+                elif len(m15_data) >= 2:
+                    prev_close = m15_data.iloc[-2]['close']
+                    if prev_close < bb_middle and m15_latest['close'] > bb_middle:
+                        buy_signals.append("突破布林中轨")
+                        signal_weights['buy'] += 1
+                    elif prev_close > bb_middle and m15_latest['close'] < bb_middle:
+                        sell_signals.append("跌破布林中轨")
+                        signal_weights['sell'] += 1
+
+        # === KDJ 信号 ===
+        if 'k' in m15_latest and 'd' in m15_latest and len(m15_data) >= 2:
+            prev = m15_data.iloc[-2]
+            if pd.notna(m15_latest['k']) and pd.notna(prev['k']):
+                if prev['k'] <= prev['d'] and m15_latest['k'] > m15_latest['d']:
+                    if m15_latest['k'] < 30:
+                        buy_signals.append("KDJ低位金叉")
+                        signal_weights['buy'] += 1.5
+                    else:
+                        buy_signals.append("KDJ金叉")
+                        signal_weights['buy'] += 0.5
+                elif prev['k'] >= prev['d'] and m15_latest['k'] < m15_latest['d']:
+                    if m15_latest['k'] > 70:
+                        sell_signals.append("KDJ高位死叉")
+                        signal_weights['sell'] += 1.5
+                    else:
+                        sell_signals.append("KDJ死叉")
+                        signal_weights['sell'] += 0.5
+
+        # === 根据趋势和阶段决定动作 ===
+        action = 'hold'
+        confidence = 0
+        reasons = []
+        signal_grade = 'D'
+
+        # 波段交易核心逻辑：在回调中寻找入场机会
+        if trend_direction == 'bullish':
+            if trend_phase == 'correction' and signal_weights['buy'] >= 3:
+                # 上涨趋势 + 回调 + 买入信号 = 最佳做多机会
+                action = 'buy'
+                confidence = min(40 + signal_weights['buy'] * 10, 95)
+                reasons = buy_signals + [f"上涨趋势回调({trend_strength})"]
+                signal_grade = 'A' if confidence >= 80 else ('B' if confidence >= 60 else 'C')
+            elif trend_phase == 'impulse' and signal_weights['buy'] >= 4:
+                # 主升浪中追多需要更强信号
+                action = 'buy'
+                confidence = min(30 + signal_weights['buy'] * 8, 80)
+                reasons = buy_signals + ["主升浪追多"]
+                signal_grade = 'B' if confidence >= 60 else 'C'
+            elif trend_phase in ['oversold', 'overbought']:
+                reasons = ['极端行情，等待企稳']
+
+        elif trend_direction == 'bearish':
+            if trend_phase == 'correction' and signal_weights['sell'] >= 3:
+                # 下跌趋势 + 反弹 + 卖出信号 = 最佳做空机会
+                action = 'sell'
+                confidence = min(40 + signal_weights['sell'] * 10, 95)
+                reasons = sell_signals + [f"下跌趋势反弹({trend_strength})"]
+                signal_grade = 'A' if confidence >= 80 else ('B' if confidence >= 60 else 'C')
+            elif trend_phase == 'impulse' and signal_weights['sell'] >= 4:
+                action = 'sell'
+                confidence = min(30 + signal_weights['sell'] * 8, 80)
+                reasons = sell_signals + ["主跌浪追空"]
+                signal_grade = 'B' if confidence >= 60 else 'C'
+            elif trend_phase in ['oversold', 'overbought']:
+                reasons = ['极端行情，等待企稳']
+
+        else:  # neutral
+            # 震荡市不交易
+            reasons = ['趋势不明确，观望']
+
+        if not reasons:
+            reasons = ['信号不足，继续观望']
+
+        # 收集指标数据
+        indicators = {}
+        for col in ['rsi', 'macd', 'macd_signal', 'macd_hist', 'k', 'd', 'j', 'close', 'ma20']:
+            if col in m15_latest and pd.notna(m15_latest[col]):
+                indicators[col] = float(m15_latest[col])
+
+        return {
+            'action': action,
+            'confidence': confidence,
+            'signal_grade': signal_grade,
+            'reasons': reasons,
+            'indicators': indicators,
+            'trend_info': {
+                'direction': trend_direction,
+                'phase': trend_phase,
+                'strength': trend_strength
+            }
+        }
+
+    async def llm_analyze(self, data: Dict[str, pd.DataFrame], signal: Dict[str, Any],
+                          symbol: str) -> Dict[str, Any]:
+        """
+        使用 LLM 进行深度分析
+
+        Args:
+            data: 多周期K线数据
+            signal: 初步信号分析结果
+            symbol: 交易对
+
+        Returns:
+            LLM 分析结果（结构化数据）
+        """
+        try:
+            # 构建分析提示
+            prompt = self._build_analysis_prompt(data, signal, symbol)
+
+            # 调用 LLM
+            response = llm_service.chat([
+                {"role": "system", "content": self.CRYPTO_ANALYST_PROMPT},
+                {"role": "user", "content": prompt}
+            ])
+
+            if response:
+                logger.info(f"LLM 分析完成: {symbol}")
+                # 解析 JSON 响应
+                return self._parse_llm_response(response)
+            else:
+                return {"error": "LLM 分析暂时不可用", "raw": ""}
+
+        except Exception as e:
+            logger.error(f"LLM 分析失败: {e}")
+            return {"error": str(e), "raw": ""}
+
+    def _parse_llm_response(self, response: str) -> Dict[str, Any]:
+        """
+        解析 LLM 的 JSON 响应
+
+        Args:
+            response: LLM 原始响应
+
+        Returns:
+            解析后的结构化数据
+        """
+        import json
+        import re
+
+        result = {
+            "raw": response,
+            "parsed": None,
+            "summary": ""
+        }
+
+        try:
+            # 尝试提取 JSON 块
+            json_match = re.search(r'```json\s*([\s\S]*?)\s*```', response)
+            if json_match:
+                json_str = json_match.group(1)
+            else:
+                # 尝试直接解析整个响应
+                json_str = response
+
+            # 解析 JSON
+            parsed = json.loads(json_str)
+            result["parsed"] = parsed
+
+            # 生成摘要
+            if parsed:
+                recommendation = parsed.get("recommendation", {})
+                action = recommendation.get("action", "wait")
+                confidence = recommendation.get("confidence", 0)
+                reason = recommendation.get("reason", "")
+
+                if action == "wait":
+                    result["summary"] = f"建议观望。{parsed.get('risk_warning', '')}"
+                else:
+                    action_text = "做多" if action == "buy" else "做空"
+                    result["summary"] = f"建议{action_text}，置信度{confidence}%。{reason}"
+
+        except json.JSONDecodeError:
+            # JSON 解析失败，提取关键信息
+            logger.warning("LLM 响应不是有效 JSON，尝试提取关键信息")
+            result["summary"] = self._extract_summary_from_text(response)
+
+        return result
+
+    def _extract_summary_from_text(self, text: str) -> str:
+        """从非 JSON 文本中提取摘要"""
+        # 简单提取前 200 字符作为摘要
+        text = text.strip()
+        if len(text) > 200:
+            return text[:200] + "..."
+        return text
+
+    def _build_analysis_prompt(self, data: Dict[str, pd.DataFrame], signal: Dict[str, Any],
+                               symbol: str) -> str:
+        """构建 LLM 分析提示 - 优化版"""
+        parts = [f"# {symbol} 技术分析数据\n"]
+
+        # 当前价格
+        current_price = float(data['5m'].iloc[-1]['close'])
+        parts.append(f"**当前价格**: ${current_price:,.2f}\n")
+
+        # 添加各周期指标摘要
+        for interval in ['4h', '1h', '15m']:
+            df = data.get(interval)
+            if df is None or df.empty:
+                continue
+
+            latest = df.iloc[-1]
+            parts.append(f"\n## {interval.upper()} 周期指标")
+
+            # 价格与均线关系
+            close = latest.get('close', 0)
+            ma20 = latest.get('ma20', 0)
+            ma50 = latest.get('ma50', 0)
+            if pd.notna(ma20):
+                position = "上方" if close > ma20 else "下方"
+                parts.append(f"- 价格在 MA20 {position} (MA20={ma20:.2f})")
+            if pd.notna(ma50):
+                parts.append(f"- MA50: {ma50:.2f}")
+
+            # RSI
+            rsi = latest.get('rsi', 0)
+            if pd.notna(rsi):
+                rsi_status = "超卖" if rsi < 30 else ("超买" if rsi > 70 else "中性")
+                parts.append(f"- RSI: {rsi:.1f} ({rsi_status})")
+
+            # MACD
+            macd = latest.get('macd', 0)
+            macd_signal = latest.get('macd_signal', 0)
+            if pd.notna(macd) and pd.notna(macd_signal):
+                macd_status = "多头" if macd > macd_signal else "空头"
+                parts.append(f"- MACD: {macd:.2f}, Signal: {macd_signal:.2f} ({macd_status})")
+
+            # 布林带
+            bb_upper = latest.get('bb_upper', 0)
+            bb_lower = latest.get('bb_lower', 0)
+            if pd.notna(bb_upper) and pd.notna(bb_lower):
+                parts.append(f"- 布林带: 上轨={bb_upper:.2f}, 下轨={bb_lower:.2f}")
+
+        # 添加最近 5 根 15M K 线（让 LLM 看形态）
+        parts.append("\n## 最近 5 根 15M K线")
+        parts.append("| 时间 | 开盘 | 最高 | 最低 | 收盘 | 涨跌 |")
+        parts.append("|------|------|------|------|------|------|")
+
+        df_15m = data.get('15m')
+        if df_15m is not None and len(df_15m) >= 5:
+            for i in range(-5, 0):
+                row = df_15m.iloc[i]
+                change = ((row['close'] - row['open']) / row['open']) * 100
+                change_str = f"+{change:.2f}%" if change >= 0 else f"{change:.2f}%"
+                time_str = row['open_time'].strftime('%H:%M') if pd.notna(row['open_time']) else 'N/A'
+                parts.append(f"| {time_str} | {row['open']:.2f} | {row['high']:.2f} | {row['low']:.2f} | {row['close']:.2f} | {change_str} |")
+
+        # 计算关键价位
+        parts.append("\n## 关键价位参考")
+        df_1h = data.get('1h')
+        if df_1h is not None and len(df_1h) >= 20:
+            recent_high = df_1h['high'].tail(20).max()
+            recent_low = df_1h['low'].tail(20).min()
+            parts.append(f"- 近期高点: ${recent_high:,.2f}")
+            parts.append(f"- 近期低点: ${recent_low:,.2f}")
+
+        # 初步信号分析结果
+        parts.append(f"\n## 规则引擎初步判断")
+        parts.append(f"- 趋势: {signal.get('trend', 'unknown')}")
+        parts.append(f"- 信号: {signal.get('action', 'hold')}")
+        parts.append(f"- 置信度: {signal.get('confidence', 0)}%")
+        parts.append(f"- 触发原因: {', '.join(signal.get('reasons', []))}")
+
+        parts.append("\n---")
+        parts.append("请基于以上数据进行分析，严格按照 JSON 格式输出你的判断。")
+
+        return "\n".join(parts)
+
+    def calculate_stop_loss_take_profit(self, price: float, action: str,
+                                        atr: float) -> Dict[str, float]:
+        """
+        计算止损止盈位置
+
+        Args:
+            price: 当前价格
+            action: 'buy' 或 'sell'
+            atr: ATR 值
+
+        Returns:
+            {'stop_loss': float, 'take_profit': float}
+        """
+        if action == 'buy':
+            stop_loss = price - atr * 2
+            take_profit = price + atr * 3
+        elif action == 'sell':
+            stop_loss = price + atr * 2
+            take_profit = price - atr * 3
+        else:
+            stop_loss = 0
+            take_profit = 0
+
+        return {
+            'stop_loss': round(stop_loss, 2),
+            'take_profit': round(take_profit, 2)
+        }
diff --git a/backend/app/crypto_agent/strategy.py b/backend/app/crypto_agent/strategy.py
new file mode 100644
index 0000000..31c3878
--- /dev/null
+++ b/backend/app/crypto_agent/strategy.py
@@ -0,0 +1,189 @@
+"""
+交易策略 - 趋势跟踪策略定义
+"""
+from typing import Dict, List, Any
+from app.utils.logger import logger
+
+
+class TrendFollowingStrategy:
+    """趋势跟踪策略"""
+
+    # 趋势判断规则（1H + 4H）
+    TREND_RULES = {
+        'bullish': {
+            'description': '看涨趋势',
+            'conditions': [
+                {'name': 'price_above_ma20', 'desc': '价格在MA20上方'},
+                {'name': 'ma5_above_ma20', 'desc': 'MA5在MA20上方'},
+                {'name': 'macd_positive', 'desc': 'MACD在信号线上方'},
+                {'name': 'rsi_above_50', 'desc': 'RSI大于50'}
+            ]
+        },
+        'bearish': {
+            'description': '看跌趋势',
+            'conditions': [
+                {'name': 'price_below_ma20', 'desc': '价格在MA20下方'},
+                {'name': 'ma5_below_ma20', 'desc': 'MA5在MA20下方'},
+                {'name': 'macd_negative', 'desc': 'MACD在信号线下方'},
+                {'name': 'rsi_below_50', 'desc': 'RSI小于50'}
+            ]
+        }
+    }
+
+    # 进场规则（5M + 15M）
+    ENTRY_RULES = {
+        'buy': {
+            'description': '做多进场',
+            'conditions': [
+                {'name': 'rsi_oversold_recovery', 'desc': 'RSI从超卖区回升', 'weight': 2},
+                {'name': 'macd_golden_cross', 'desc': 'MACD金叉', 'weight': 2},
+                {'name': 'price_break_bb_middle', 'desc': '价格突破布林中轨', 'weight': 1},
+                {'name': 'kdj_golden_cross', 'desc': 'KDJ低位金叉', 'weight': 1},
+                {'name': 'volume_increase', 'desc': '成交量放大', 'weight': 1}
+            ],
+            'min_score': 3  # 最低触发分数
+        },
+        'sell': {
+            'description': '做空进场',
+            'conditions': [
+                {'name': 'rsi_overbought_decline', 'desc': 'RSI从超买区回落', 'weight': 2},
+                {'name': 'macd_death_cross', 'desc': 'MACD死叉', 'weight': 2},
+                {'name': 'price_break_bb_middle_down', 'desc': '价格跌破布林中轨', 'weight': 1},
+                {'name': 'kdj_death_cross', 'desc': 'KDJ高位死叉', 'weight': 1},
+                {'name': 'volume_increase', 'desc': '成交量放大', 'weight': 1}
+            ],
+            'min_score': 3
+        }
+    }
+
+    # 出场规则
+    EXIT_RULES = {
+        'take_profit': {
+            'description': '止盈',
+            'conditions': [
+                {'name': 'target_reached', 'desc': '达到目标价位'},
+                {'name': 'rsi_extreme', 'desc': 'RSI达到极值'},
+                {'name': 'trend_reversal', 'desc': '趋势反转信号'}
+            ]
+        },
+        'stop_loss': {
+            'description': '止损',
+            'conditions': [
+                {'name': 'price_hit_stop', 'desc': '价格触及止损位'},
+                {'name': 'trend_break', 'desc': '趋势破坏'}
+            ]
+        }
+    }
+
+    # 风险管理参数
+    RISK_PARAMS = {
+        'max_position_size': 0.1,  # 最大仓位比例
+        'stop_loss_atr_multiplier': 2.0,  # 止损 ATR 倍数
+        'take_profit_atr_multiplier': 3.0,  # 止盈 ATR 倍数
+        'max_daily_trades': 5,  # 每日最大交易次数
+        'min_risk_reward_ratio': 1.5  # 最小风险收益比
+    }
+
+    def __init__(self):
+        """初始化策略"""
+        logger.info("趋势跟踪策略初始化完成")
+
+    def get_trend_rules(self, trend: str) -> Dict[str, Any]:
+        """获取趋势判断规则"""
+        return self.TREND_RULES.get(trend, {})
+
+    def get_entry_rules(self, action: str) -> Dict[str, Any]:
+        """获取进场规则"""
+        return self.ENTRY_RULES.get(action, {})
+
+    def get_exit_rules(self, exit_type: str) -> Dict[str, Any]:
+        """获取出场规则"""
+        return self.EXIT_RULES.get(exit_type, {})
+
+    def calculate_position_size(self, account_balance: float, risk_per_trade: float,
+                                entry_price: float, stop_loss: float) -> float:
+        """
+        计算仓位大小
+
+        Args:
+            account_balance: 账户余额
+            risk_per_trade: 单笔风险比例（如 0.02 表示 2%）
+            entry_price: 入场价格
+            stop_loss: 止损价格
+
+        Returns:
+            建议仓位大小
+        """
+        risk_amount = account_balance * risk_per_trade
+        price_risk = abs(entry_price - stop_loss)
+
+        if price_risk == 0:
+            return 0
+
+        position_size = risk_amount / price_risk
+
+        # 限制最大仓位
+        max_position = account_balance * self.RISK_PARAMS['max_position_size'] / entry_price
+        position_size = min(position_size, max_position)
+
+        return round(position_size, 6)
+
+    def validate_trade(self, entry_price: float, stop_loss: float,
+                       take_profit: float) -> Dict[str, Any]:
+        """
+        验证交易是否符合风险管理规则
+
+        Args:
+            entry_price: 入场价格
+            stop_loss: 止损价格
+            take_profit: 止盈价格
+
+        Returns:
+            验证结果
+        """
+        risk = abs(entry_price - stop_loss)
+        reward = abs(take_profit - entry_price)
+
+        if risk == 0:
+            return {
+                'valid': False,
+                'reason': '止损距离为0',
+                'risk_reward_ratio': 0
+            }
+
+        risk_reward_ratio = reward / risk
+
+        if risk_reward_ratio < self.RISK_PARAMS['min_risk_reward_ratio']:
+            return {
+                'valid': False,
+                'reason': f'风险收益比({risk_reward_ratio:.2f})低于最低要求({self.RISK_PARAMS["min_risk_reward_ratio"]})',
+                'risk_reward_ratio': risk_reward_ratio
+            }
+
+        return {
+            'valid': True,
+            'reason': '符合风险管理规则',
+            'risk_reward_ratio': risk_reward_ratio
+        }
+
+    def get_strategy_description(self) -> str:
+        """获取策略描述"""
+        return """
+## 趋势跟踪策略
+
+### 核心理念
+顺势而为，在大周期确认趋势后，在小周期寻找最佳进场点。
+
+### 趋势判断（4H + 1H）
+- 看涨：价格在MA20上方，MA5>MA20，MACD>信号线，RSI>50
+- 看跌：价格在MA20下方，MA5<MA20，MACD<信号线，RSI<50
+
+### 进场信号（15M + 5M）
+- 做多：RSI超卖回升、MACD金叉、突破布林中轨、KDJ低位金叉
+- 做空：RSI超买回落、MACD死叉、跌破布林中轨、KDJ高位死叉
+
+### 风险管理
+- 止损：2倍ATR
+- 止盈：3倍ATR
+- 最小风险收益比：1.5
+"""
diff --git a/backend/app/services/binance_service.py b/backend/app/services/binance_service.py
new file mode 100644
index 0000000..0dbf3dc
--- /dev/null
+++ b/backend/app/services/binance_service.py
@@ -0,0 +1,246 @@
+"""
+Binance 数据服务 - 获取加密货币 K 线数据和技术指标
+"""
+import pandas as pd
+import numpy as np
+from typing import Dict, List, Optional, Any
+from binance.client import Client
+from binance.enums import (
+    KLINE_INTERVAL_5MINUTE,
+    KLINE_INTERVAL_15MINUTE,
+    KLINE_INTERVAL_1HOUR,
+    KLINE_INTERVAL_4HOUR
+)
+from app.utils.logger import logger
+
+
+class BinanceService:
+    """Binance 数据服务"""
+
+    # K线周期映射
+    INTERVALS = {
+        '5m': KLINE_INTERVAL_5MINUTE,
+        '15m': KLINE_INTERVAL_15MINUTE,
+        '1h': KLINE_INTERVAL_1HOUR,
+        '4h': KLINE_INTERVAL_4HOUR
+    }
+
+    def __init__(self, api_key: str = "", api_secret: str = ""):
+        """
+        初始化 Binance 客户端
+
+        Args:
+            api_key: API 密钥（可选，公开数据不需要）
+            api_secret: API 密钥（可选）
+        """
+        self.client = Client(api_key=api_key, api_secret=api_secret)
+        logger.info("Binance 服务初始化完成")
+
+    def get_klines(self, symbol: str, interval: str, limit: int = 100) -> pd.DataFrame:
+        """
+        获取 K 线数据
+
+        Args:
+            symbol: 交易对，如 'BTCUSDT'
+            interval: K线周期，如 '5m', '15m', '1h', '4h'
+            limit: 获取数量
+
+        Returns:
+            DataFrame 包含 OHLCV 数据
+        """
+        try:
+            binance_interval = self.INTERVALS.get(interval, interval)
+            klines = self.client.get_klines(
+                symbol=symbol,
+                interval=binance_interval,
+                limit=limit
+            )
+            return self._parse_klines(klines)
+        except Exception as e:
+            logger.error(f"获取 {symbol} {interval} K线数据失败: {e}")
+            return pd.DataFrame()
+
+    def get_multi_timeframe_data(self, symbol: str) -> Dict[str, pd.DataFrame]:
+        """
+        获取多周期 K 线数据
+
+        Args:
+            symbol: 交易对
+
+        Returns:
+            包含 5m, 15m, 1h, 4h 数据的字典
+        """
+        data = {}
+        for interval in ['5m', '15m', '1h', '4h']:
+            df = self.get_klines(symbol, interval, limit=100)
+            if not df.empty:
+                df = self.calculate_indicators(df)
+            data[interval] = df
+
+        logger.info(f"获取 {symbol} 多周期数据完成")
+        return data
+
+    def _parse_klines(self, klines: List) -> pd.DataFrame:
+        """解析 K 线数据为 DataFrame"""
+        if not klines:
+            return pd.DataFrame()
+
+        df = pd.DataFrame(klines, columns=[
+            'open_time', 'open', 'high', 'low', 'close', 'volume',
+            'close_time', 'quote_volume', 'trades',
+            'taker_buy_base', 'taker_buy_quote', 'ignore'
+        ])
+
+        # 转换数据类型
+        df['open_time'] = pd.to_datetime(df['open_time'], unit='ms')
+        df['close_time'] = pd.to_datetime(df['close_time'], unit='ms')
+
+        for col in ['open', 'high', 'low', 'close', 'volume', 'quote_volume']:
+            df[col] = df[col].astype(float)
+
+        df['trades'] = df['trades'].astype(int)
+
+        # 只保留需要的列
+        df = df[['open_time', 'open', 'high', 'low', 'close', 'volume', 'trades']]
+
+        return df
+
+    def calculate_indicators(self, df: pd.DataFrame) -> pd.DataFrame:
+        """
+        计算技术指标
+
+        Args:
+            df: K线数据 DataFrame
+
+        Returns:
+            添加了技术指标的 DataFrame
+        """
+        if df.empty:
+            return df
+
+        # 移动平均线
+        df['ma5'] = self._calculate_ma(df['close'], 5)
+        df['ma10'] = self._calculate_ma(df['close'], 10)
+        df['ma20'] = self._calculate_ma(df['close'], 20)
+        df['ma50'] = self._calculate_ma(df['close'], 50)
+
+        # EMA
+        df['ema12'] = self._calculate_ema(df['close'], 12)
+        df['ema26'] = self._calculate_ema(df['close'], 26)
+
+        # RSI
+        df['rsi'] = self._calculate_rsi(df['close'], 14)
+
+        # MACD
+        df['macd'], df['macd_signal'], df['macd_hist'] = self._calculate_macd(df['close'])
+
+        # 布林带
+        df['bb_upper'], df['bb_middle'], df['bb_lower'] = self._calculate_bollinger(df['close'])
+
+        # KDJ
+        df['k'], df['d'], df['j'] = self._calculate_kdj(df['high'], df['low'], df['close'])
+
+        # ATR
+        df['atr'] = self._calculate_atr(df['high'], df['low'], df['close'])
+
+        return df
+
+    @staticmethod
+    def _calculate_ma(data: pd.Series, period: int) -> pd.Series:
+        """简单移动平均线"""
+        return data.rolling(window=period).mean()
+
+    @staticmethod
+    def _calculate_ema(data: pd.Series, period: int) -> pd.Series:
+        """指数移动平均线"""
+        return data.ewm(span=period, adjust=False).mean()
+
+    @staticmethod
+    def _calculate_rsi(data: pd.Series, period: int = 14) -> pd.Series:
+        """RSI 指标"""
+        delta = data.diff()
+        gain = (delta.where(delta > 0, 0)).rolling(window=period).mean()
+        loss = (-delta.where(delta < 0, 0)).rolling(window=period).mean()
+        rs = gain / loss
+        rsi = 100 - (100 / (1 + rs))
+        return rsi
+
+    @staticmethod
+    def _calculate_macd(data: pd.Series, fast: int = 12, slow: int = 26, signal: int = 9):
+        """MACD 指标"""
+        ema_fast = data.ewm(span=fast, adjust=False).mean()
+        ema_slow = data.ewm(span=slow, adjust=False).mean()
+
+        macd = ema_fast - ema_slow
+        signal_line = macd.ewm(span=signal, adjust=False).mean()
+        histogram = macd - signal_line
+
+        return macd, signal_line, histogram
+
+    @staticmethod
+    def _calculate_bollinger(data: pd.Series, period: int = 20, std_dev: float = 2.0):
+        """布林带"""
+        middle = data.rolling(window=period).mean()
+        std = data.rolling(window=period).std()
+
+        upper = middle + (std * std_dev)
+        lower = middle - (std * std_dev)
+
+        return upper, middle, lower
+
+    @staticmethod
+    def _calculate_kdj(high: pd.Series, low: pd.Series, close: pd.Series,
+                       period: int = 9, k_period: int = 3, d_period: int = 3):
+        """KDJ 指标"""
+        low_min = low.rolling(window=period).min()
+        high_max = high.rolling(window=period).max()
+
+        rsv = (close - low_min) / (high_max - low_min) * 100
+
+        k = rsv.ewm(com=k_period - 1, adjust=False).mean()
+        d = k.ewm(com=d_period - 1, adjust=False).mean()
+        j = 3 * k - 2 * d
+
+        return k, d, j
+
+    @staticmethod
+    def _calculate_atr(high: pd.Series, low: pd.Series, close: pd.Series, period: int = 14):
+        """ATR 平均真实波幅"""
+        tr1 = high - low
+        tr2 = abs(high - close.shift())
+        tr3 = abs(low - close.shift())
+
+        tr = pd.concat([tr1, tr2, tr3], axis=1).max(axis=1)
+        atr = tr.rolling(window=period).mean()
+
+        return atr
+
+    def get_current_price(self, symbol: str) -> Optional[float]:
+        """获取当前价格"""
+        try:
+            ticker = self.client.get_symbol_ticker(symbol=symbol)
+            return float(ticker['price'])
+        except Exception as e:
+            logger.error(f"获取 {symbol} 当前价格失败: {e}")
+            return None
+
+    def get_24h_stats(self, symbol: str) -> Optional[Dict[str, Any]]:
+        """获取 24 小时统计数据"""
+        try:
+            stats = self.client.get_ticker(symbol=symbol)
+            return {
+                'price': float(stats['lastPrice']),
+                'price_change': float(stats['priceChange']),
+                'price_change_percent': float(stats['priceChangePercent']),
+                'high': float(stats['highPrice']),
+                'low': float(stats['lowPrice']),
+                'volume': float(stats['volume']),
+                'quote_volume': float(stats['quoteVolume'])
+            }
+        except Exception as e:
+            logger.error(f"获取 {symbol} 24h 统计失败: {e}")
+            return None
+
+
+# 全局实例
+binance_service = BinanceService()
diff --git a/backend/app/services/feishu_service.py b/backend/app/services/feishu_service.py
new file mode 100644
index 0000000..e4a578b
--- /dev/null
+++ b/backend/app/services/feishu_service.py
@@ -0,0 +1,296 @@
+"""
+飞书通知服务 - 通过 Webhook 发送交易信号通知
+"""
+import json
+import httpx
+from typing import Dict, Any, Optional
+from app.utils.logger import logger
+from app.config import get_settings
+
+
+class FeishuService:
+    """飞书机器人通知服务"""
+
+    def __init__(self, webhook_url: str = ""):
+        """
+        初始化飞书服务
+
+        Args:
+            webhook_url: 飞书机器人 Webhook URL
+        """
+        settings = get_settings()
+        self.webhook_url = webhook_url or getattr(settings, 'feishu_webhook_url', '')
+        self.enabled = bool(self.webhook_url)
+
+        if self.enabled:
+            logger.info("飞书通知服务初始化完成")
+        else:
+            logger.warning("飞书 Webhook URL 未配置，通知功能已禁用")
+
+    async def send_text(self, message: str) -> bool:
+        """
+        发送文本消息
+
+        Args:
+            message: 消息内容
+
+        Returns:
+            是否发送成功
+        """
+        if not self.enabled:
+            logger.warning("飞书服务未启用，跳过发送")
+            return False
+
+        data = {
+            "msg_type": "text",
+            "content": {
+                "text": message
+            }
+        }
+
+        return await self._send(data)
+
+    async def send_card(self, title: str, content: str, color: str = "blue") -> bool:
+        """
+        发送卡片消息
+
+        Args:
+            title: 卡片标题
+            content: 卡片内容（支持 Markdown）
+            color: 标题颜色 (blue, green, red, orange, purple)
+
+        Returns:
+            是否发送成功
+        """
+        if not self.enabled:
+            logger.warning("飞书服务未启用，跳过发送")
+            return False
+
+        # 颜色映射
+        color_map = {
+            "blue": "blue",
+            "green": "green",
+            "red": "red",
+            "orange": "orange",
+            "purple": "purple"
+        }
+
+        data = {
+            "msg_type": "interactive",
+            "card": {
+                "header": {
+                    "title": {
+                        "tag": "plain_text",
+                        "content": title
+                    },
+                    "template": color_map.get(color, "blue")
+                },
+                "elements": [
+                    {
+                        "tag": "markdown",
+                        "content": content
+                    }
+                ]
+            }
+        }
+
+        return await self._send(data)
+
+    async def send_trading_signal(self, signal: Dict[str, Any]) -> bool:
+        """
+        发送交易信号卡片
+
+        Args:
+            signal: 交易信号数据
+                - symbol: 交易对
+                - action: 'buy' | 'sell'
+                - price: 当前价格
+                - trend: 趋势方向
+                - confidence: 信号强度 (0-100)
+                - indicators: 技术指标数据
+                - llm_analysis: LLM 分析结果（可选）
+                - stop_loss: 建议止损价
+                - take_profit: 建议止盈价
+
+        Returns:
+            是否发送成功
+        """
+        if not self.enabled:
+            logger.warning("飞书服务未启用，跳过发送")
+            return False
+
+        action = signal.get('action', 'hold')
+        symbol = signal.get('symbol', 'UNKNOWN')
+        price = signal.get('price', 0)
+        trend = signal.get('trend', 'neutral')
+        confidence = signal.get('confidence', 0)
+        indicators = signal.get('indicators', {})
+        llm_analysis = signal.get('llm_analysis', '')
+        stop_loss = signal.get('stop_loss', 0)
+        take_profit = signal.get('take_profit', 0)
+
+        # 确定标题和颜色
+        if action == 'buy':
+            title = f"🟢 买入信号 - {symbol}"
+            color = "green"
+            action_text = "做多"
+        elif action == 'sell':
+            title = f"🔴 卖出信号 - {symbol}"
+            color = "red"
+            action_text = "做空"
+        else:
+            title = f"⚪ 观望 - {symbol}"
+            color = "blue"
+            action_text = "观望"
+
+        # 趋势文本
+        trend_text = {
+            'bullish': '看涨 📈',
+            'bearish': '看跌 📉',
+            'neutral': '震荡 ↔️'
+        }.get(trend, '未知')
+
+        # 构建内容
+        content_parts = [
+            f"**当前价格**: ${price:,.2f}",
+            f"**趋势方向**: {trend_text}",
+            f"**信号强度**: {confidence}%",
+            "",
+            "---",
+            "",
+            "**技术指标**:"
+        ]
+
+        # 添加技术指标
+        if indicators:
+            rsi = indicators.get('rsi', 0)
+            macd = indicators.get('macd', 0)
+            macd_signal = indicators.get('macd_signal', 0)
+
+            rsi_status = "超卖 ↑" if rsi < 30 else ("超买 ↓" if rsi > 70 else "中性")
+            macd_status = "金叉" if macd > macd_signal else "死叉"
+
+            content_parts.extend([
+                f"• RSI(14): {rsi:.1f} ({rsi_status})",
+                f"• MACD: {macd_status}",
+            ])
+
+            if 'k' in indicators:
+                content_parts.append(f"• KDJ: K={indicators['k']:.1f}, D={indicators['d']:.1f}")
+
+        # 添加 LLM 分析
+        if llm_analysis:
+            content_parts.extend([
+                "",
+                "---",
+                "",
+                "**AI 分析**:",
+                llm_analysis[:200] + "..." if len(llm_analysis) > 200 else llm_analysis
+            ])
+
+        # 添加止损止盈建议
+        if stop_loss > 0 or take_profit > 0:
+            content_parts.extend([
+                "",
+                "---",
+                "",
+                "**风险管理**:"
+            ])
+            if stop_loss > 0:
+                sl_percent = ((stop_loss - price) / price) * 100
+                content_parts.append(f"• 建议止损: ${stop_loss:,.2f} ({sl_percent:+.1f}%)")
+            if take_profit > 0:
+                tp_percent = ((take_profit - price) / price) * 100
+                content_parts.append(f"• 建议止盈: ${take_profit:,.2f} ({tp_percent:+.1f}%)")
+
+        # 添加免责声明
+        content_parts.extend([
+            "",
+            "---",
+            "",
+            "*⚠️ 仅供参考，不构成投资建议*"
+        ])
+
+        content = "\n".join(content_parts)
+
+        return await self.send_card(title, content, color)
+
+    async def send_trend_change(self, symbol: str, old_trend: str, new_trend: str, price: float) -> bool:
+        """
+        发送趋势变化通知
+
+        Args:
+            symbol: 交易对
+            old_trend: 旧趋势
+            new_trend: 新趋势
+            price: 当前价格
+
+        Returns:
+            是否发送成功
+        """
+        trend_emoji = {
+            'bullish': '📈',
+            'bearish': '📉',
+            'neutral': '↔️'
+        }
+
+        trend_text = {
+            'bullish': '看涨',
+            'bearish': '看跌',
+            'neutral': '震荡'
+        }
+
+        title = f"🔄 趋势变化 - {symbol}"
+        content = f"""**{symbol}** 趋势发生变化
+
+**变化**: {trend_text.get(old_trend, old_trend)} {trend_emoji.get(old_trend, '')} → {trend_text.get(new_trend, new_trend)} {trend_emoji.get(new_trend, '')}
+
+**当前价格**: ${price:,.2f}
+
+*请关注后续交易信号*"""
+
+        return await self.send_card(title, content, "orange")
+
+    async def _send(self, data: Dict[str, Any]) -> bool:
+        """
+        发送消息到飞书
+
+        Args:
+            data: 消息数据
+
+        Returns:
+            是否发送成功
+        """
+        try:
+            async with httpx.AsyncClient() as client:
+                response = await client.post(
+                    self.webhook_url,
+                    json=data,
+                    headers={"Content-Type": "application/json"},
+                    timeout=10.0
+                )
+
+                result = response.json()
+
+                if result.get('code') == 0 or result.get('StatusCode') == 0:
+                    logger.info("飞书消息发送成功")
+                    return True
+                else:
+                    logger.error(f"飞书消息发送失败: {result}")
+                    return False
+
+        except Exception as e:
+            logger.error(f"飞书消息发送异常: {e}")
+            return False
+
+
+# 全局实例（延迟初始化）
+_feishu_service: Optional[FeishuService] = None
+
+
+def get_feishu_service() -> FeishuService:
+    """获取飞书服务实例"""
+    global _feishu_service
+    if _feishu_service is None:
+        _feishu_service = FeishuService()
+    return _feishu_service
diff --git a/backend/requirements.txt b/backend/requirements.txt
index bba367f..7f06878 100644
--- a/backend/requirements.txt
+++ b/backend/requirements.txt
@@ -19,3 +19,7 @@ yfinance>=0.2.36
 PyJWT==2.8.0
 tencentcloud-sdk-python==3.0.1100
 python-jose[cryptography]==3.3.0
+
+# 加密货币交易智能体依赖
+python-binance>=1.0.19
+httpx>=0.27.0
diff --git a/backend/run_crypto.sh b/backend/run_crypto.sh
new file mode 100755
index 0000000..a6ad8f1
--- /dev/null
+++ b/backend/run_crypto.sh
@@ -0,0 +1,85 @@
+#!/bin/bash
+
+# 加密货币智能体启动脚本
+
+echo "================================"
+echo "加密货币交易信号智能体"
+echo "================================"
+echo ""
+
+cd /Users/aaron/source_code/Stock_Agent/backend
+
+# 激活虚拟环境
+if [ ! -d "venv" ]; then
+    echo "❌ 虚拟环境不存在，请先运行 ../install.sh"
+    exit 1
+fi
+
+source venv/bin/activate
+
+# 检查依赖
+echo "1. 检查依赖..."
+python3 -c "import binance; print('   ✓ python-binance')" 2>/dev/null || {
+    echo "   安装 python-binance..."
+    pip install python-binance -q
+}
+python3 -c "import httpx; print('   ✓ httpx')" 2>/dev/null || {
+    echo "   安装 httpx..."
+    pip install httpx -q
+}
+
+# 测试导入
+echo ""
+echo "2. 测试模块导入..."
+python3 << 'EOF'
+try:
+    from app.services.binance_service import binance_service
+    print("   ✓ Binance 服务")
+
+    from app.services.feishu_service import get_feishu_service
+    print("   ✓ 飞书服务")
+
+    from app.crypto_agent.crypto_agent import crypto_agent
+    print("   ✓ 加密货币智能体")
+
+    print("\n所有模块导入成功！")
+
+except Exception as e:
+    print(f"\n❌ 导入失败: {e}")
+    import traceback
+    traceback.print_exc()
+    exit(1)
+EOF
+
+if [ $? -ne 0 ]; then
+    echo ""
+    echo "模块导入失败，请检查错误信息"
+    exit 1
+fi
+
+# 检查配置
+echo ""
+echo "3. 检查配置..."
+python3 << 'EOF'
+from app.config import get_settings
+settings = get_settings()
+
+print(f"   监控交易对: {settings.crypto_symbols}")
+print(f"   分析间隔: {settings.crypto_analysis_interval}秒")
+print(f"   飞书 Webhook: {'✓ 已配置' if settings.feishu_webhook_url else '❌ 未配置'}")
+print(f"   LLM 服务: {'✓ 已配置' if settings.zhipuai_api_key or settings.deepseek_api_key else '⚠️ 未配置'}")
+
+if not settings.feishu_webhook_url:
+    print("\n⚠️  警告: 飞书 Webhook 未配置，将无法发送通知")
+EOF
+
+echo ""
+echo "================================"
+echo "启动智能体..."
+echo "================================"
+echo ""
+echo "按 Ctrl+C 停止"
+echo ""
+
+# 启动智能体
+python3 run_crypto_agent.py
diff --git a/backend/run_crypto_agent.py b/backend/run_crypto_agent.py
new file mode 100644
index 0000000..3cc3aaa
--- /dev/null
+++ b/backend/run_crypto_agent.py
@@ -0,0 +1,33 @@
+#!/usr/bin/env python3
+"""
+加密货币智能体启动脚本
+"""
+import asyncio
+import sys
+import os
+
+# 添加项目路径
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from app.crypto_agent.crypto_agent import crypto_agent
+from app.utils.logger import logger
+
+
+async def main():
+    """主函数"""
+    logger.info("=" * 50)
+    logger.info("加密货币交易信号智能体")
+    logger.info("=" * 50)
+
+    try:
+        await crypto_agent.run()
+    except KeyboardInterrupt:
+        logger.info("收到停止信号，正在关闭...")
+        crypto_agent.stop()
+    except Exception as e:
+        logger.error(f"运行出错: {e}")
+        raise
+
+
+if __name__ == "__main__":
+    asyncio.run(main())
diff --git a/backend/test_crypto_agent.py b/backend/test_crypto_agent.py
new file mode 100644
index 0000000..c42e58a
--- /dev/null
+++ b/backend/test_crypto_agent.py
@@ -0,0 +1,170 @@
+#!/usr/bin/env python3
+"""
+测试加密货币智能体 - 单次分析
+"""
+import asyncio
+import sys
+import os
+
+# 添加项目路径
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+
+async def test_binance_service():
+    """测试 Binance 数据服务"""
+    print("\n" + "=" * 50)
+    print("测试 Binance 数据服务")
+    print("=" * 50)
+
+    from app.services.binance_service import binance_service
+
+    # 测试获取 K 线数据
+    print("\n1. 获取 BTCUSDT 1小时 K线数据...")
+    df = binance_service.get_klines('BTCUSDT', '1h', limit=10)
+    if not df.empty:
+        print(f"   ✓ 获取成功，共 {len(df)} 条数据")
+        print(f"   最新收盘价: ${df.iloc[-1]['close']:,.2f}")
+    else:
+        print("   ✗ 获取失败")
+        return False
+
+    # 测试多周期数据
+    print("\n2. 获取 BTCUSDT 多周期数据...")
+    data = binance_service.get_multi_timeframe_data('BTCUSDT')
+    for interval, df in data.items():
+        if not df.empty:
+            print(f"   ✓ {interval}: {len(df)} 条数据")
+        else:
+            print(f"   ✗ {interval}: 无数据")
+
+    # 测试技术指标
+    print("\n3. 检查技术指标...")
+    df = data['1h']
+    indicators = ['ma5', 'ma20', 'rsi', 'macd', 'bb_upper', 'k', 'd', 'atr']
+    for ind in indicators:
+        if ind in df.columns:
+            value = df.iloc[-1][ind]
+            print(f"   ✓ {ind}: {value:.4f}" if value else f"   - {ind}: N/A")
+
+    return True
+
+
+async def test_signal_analyzer():
+    """测试信号分析器"""
+    print("\n" + "=" * 50)
+    print("测试信号分析器")
+    print("=" * 50)
+
+    from app.services.binance_service import binance_service
+    from app.crypto_agent.signal_analyzer import SignalAnalyzer
+
+    analyzer = SignalAnalyzer()
+
+    # 获取数据
+    data = binance_service.get_multi_timeframe_data('BTCUSDT')
+
+    # 测试趋势分析
+    print("\n1. 分析趋势...")
+    trend = analyzer.analyze_trend(data['1h'], data['4h'])
+    print(f"   趋势: {trend}")
+
+    # 测试进场信号
+    print("\n2. 分析进场信号...")
+    signal = analyzer.analyze_entry_signal(data['5m'], data['15m'], trend)
+    print(f"   动作: {signal['action']}")
+    print(f"   置信度: {signal['confidence']}%")
+    print(f"   原因: {', '.join(signal['reasons'])}")
+
+    return True
+
+
+async def test_feishu_service():
+    """测试飞书通知服务"""
+    print("\n" + "=" * 50)
+    print("测试飞书通知服务")
+    print("=" * 50)
+
+    from app.services.feishu_service import get_feishu_service
+
+    feishu = get_feishu_service()
+
+    if not feishu.enabled:
+        print("   ⚠ 飞书服务未配置，跳过测试")
+        return True
+
+    # 发送测试消息
+    print("\n1. 发送测试文本消息...")
+    result = await feishu.send_text("🧪 这是一条测试消息，来自加密货币智能体")
+    print(f"   {'✓ 发送成功' if result else '✗ 发送失败'}")
+
+    return result
+
+
+async def test_full_analysis():
+    """测试完整分析流程"""
+    print("\n" + "=" * 50)
+    print("测试完整分析流程")
+    print("=" * 50)
+
+    from app.crypto_agent.crypto_agent import CryptoAgent
+
+    agent = CryptoAgent()
+
+    for symbol in ['BTCUSDT', 'ETHUSDT']:
+        print(f"\n分析 {symbol}...")
+        result = await agent.analyze_once(symbol)
+
+        if 'error' in result:
+            print(f"   ✗ 错误: {result['error']}")
+        else:
+            print(f"   价格: ${result['price']:,.2f}")
+            print(f"   趋势: {result['trend']}")
+            print(f"   动作: {result['action']}")
+            print(f"   置信度: {result['confidence']}%")
+            if result.get('stop_loss'):
+                print(f"   止损: ${result['stop_loss']:,.2f}")
+            if result.get('take_profit'):
+                print(f"   止盈: ${result['take_profit']:,.2f}")
+
+    return True
+
+
+async def main():
+    """主测试函数"""
+    print("\n" + "=" * 60)
+    print("加密货币智能体测试")
+    print("=" * 60)
+
+    tests = [
+        ("Binance 数据服务", test_binance_service),
+        ("信号分析器", test_signal_analyzer),
+        ("飞书通知服务", test_feishu_service),
+        ("完整分析流程", test_full_analysis),
+    ]
+
+    results = []
+    for name, test_func in tests:
+        try:
+            result = await test_func()
+            results.append((name, result))
+        except Exception as e:
+            print(f"\n✗ {name} 测试出错: {e}")
+            results.append((name, False))
+
+    # 打印总结
+    print("\n" + "=" * 60)
+    print("测试总结")
+    print("=" * 60)
+    for name, result in results:
+        status = "✓ 通过" if result else "✗ 失败"
+        print(f"   {name}: {status}")
+
+    all_passed = all(r for _, r in results)
+    print("\n" + ("所有测试通过！" if all_passed else "部分测试失败"))
+
+    return all_passed
+
+
+if __name__ == "__main__":
+    success = asyncio.run(main())
+    sys.exit(0 if success else 1)