diff --git a/docs/discovery-workflow.md b/docs/discovery-workflow.md
index d6a6224..4508f96 100644
--- a/docs/discovery-workflow.md
+++ b/docs/discovery-workflow.md
@@ -15,14 +15,22 @@
    ├─ 筛选符合条件的项目链接
    │
    ▼
-2. 调用创建任务 API
+2. 调用去重检查 API (推荐)
+   POST /api/discovery/check-duplicates
+   │
+   ├─ 检查 URL 是否已存在任务
+   ├─ 检查 URL 对应项目是否已收录
+   ├─ 过滤出需要创建的 URL
+   │
+   ▼
+3. 调用创建任务 API
    POST /api/discovery/tasks
    │
    ├─ 存入 ProjectDiscoveryTask 表
    ├─ 状态: PENDING
    │
    ▼
-3. 手动触发本地命令 (定期执行)
+4. 手动触发本地命令 (定期执行)
    /discover-projects
    │
    ├─ 通过 curl 获取待处理任务
@@ -38,7 +46,7 @@
    │   └─ 自动重试失败的提交
    │
    ▼
-4. 完成任务并入库
+5. 完成任务并入库
    POST /api/discovery/tasks/{id}/complete
    │
    ├─ 验证数据格式 (Zod Schema)
@@ -50,7 +58,7 @@
    ├─ 更新任务状态: COMPLETED/FAILED
    │
    ▼
-5. 数据已入库，可在前台展示
+6. 数据已入库，可在前台展示
 ```
 
 ---
@@ -108,6 +116,109 @@ curl -X POST https://your-domain.com/api/discovery/tasks \
 
 ---
 
+### 步骤 2.5: 调用去重检查 API（推荐）
+
+**API 端点**: `POST /api/discovery/check-duplicates`
+
+**调用示例**:
+```bash
+curl -X POST https://your-domain.com/api/discovery/check-duplicates \
+  -H "Content-Type: application/json" \
+  -d '{
+    "apiKey": "YOUR_API_KEY",
+    "urls": [
+      "https://github.com/langchain-ai/langchain",
+      "https://github.com/openai/openai-quickstart-python"
+    ]
+  }'
+```
+
+**返回结果**:
+```json
+{
+  "success": true,
+  "results": [
+    {
+      "url": "https://github.com/langchain-ai/langchain",
+      "shouldCreate": false,
+      "reason": "Task already exists with status PENDING",
+      "existingTask": {
+        "id": "cmxxxxx",
+        "status": "PENDING",
+        "sourceUrl": "https://github.com/langchain-ai/langchain",
+        "createdAt": "2025-01-18T10:00:00Z",
+        "projectId": null
+      }
+    },
+    {
+      "url": "https://github.com/openai/openai-quickstart-python",
+      "shouldCreate": true,
+      "reason": "No existing task or project found"
+    }
+  ],
+  "stats": {
+    "total": 2,
+    "shouldCreate": 1,
+    "duplicate": 1
+  }
+}
+```
+
+**去重逻辑**（按优先级）:
+1. **优先级 1**: 检查是否有 PENDING/IN_PROGRESS 的相同 URL 任务
+   - 如果存在 → `shouldCreate: false`
+   - 原因：任务已在处理中，避免重复探索
+
+2. **优先级 2**: 检查是否有 COMPLETED/FAILED 的相同 URL 任务
+   - 如果存在 → `shouldCreate: false`
+   - 原因：任务已探索过，无需重复
+
+3. **优先级 3**: 检查 URL 对应的项目是否已存在（通过 ExternalLink）
+   - 如果存在 → `shouldCreate: false`
+   - 原因：项目已通过其他来源收录
+
+4. **默认**: 允许创建新任务
+   - `shouldCreate: true`
+
+**n8n 集成建议**:
+```javascript
+// n8n Workflow 示例
+const checkResponse = await fetch('https://your-domain.com/api/discovery/check-duplicates', {
+  method: 'POST',
+  headers: { 'Content-Type': 'application/json' },
+  body: JSON.stringify({
+    apiKey: 'YOUR_API_KEY',
+    urls: collectedUrls  // 从上一步收集的 URL 列表
+  })
+})
+
+const { results, stats } = await checkResponse.json()
+
+// 过滤出应该创建任务的 URL
+const urlsToCreate = results
+  .filter(r => r.shouldCreate)
+  .map(r => r.url)
+
+// 只为不重复的 URL 创建任务
+if (urlsToCreate.length > 0) {
+  await fetch('https://your-domain.com/api/discovery/tasks', {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify({
+      apiKey: 'YOUR_API_KEY',
+      tasks: urlsToCreate.map(url => ({
+        sourceUrl: url,
+        sourceType: 'github_trending'
+      }))
+    })
+  })
+}
+
+console.log(`创建 ${urlsToCreate.length} 个新任务，跳过 ${stats.duplicate} 个重复任务`)
+```
+
+---
+
 ### 步骤 3: 手动触发本地命令
 
 **执行环境**: 本地开发环境 (Local Development)
diff --git a/src/app/api/discovery/check-duplicates/route.ts b/src/app/api/discovery/check-duplicates/route.ts
new file mode 100644
index 0000000..3f22956
--- /dev/null
+++ b/src/app/api/discovery/check-duplicates/route.ts
@@ -0,0 +1,247 @@
+import { NextRequest, NextResponse } from 'next/server'
+import crypto from 'crypto'
+import { prisma } from '@/lib/prisma'
+import { CheckTaskDuplicatesSchema } from '@/lib/validations'
+
+/**
+ * 检查 URL 是否应该创建新任务
+ *
+ * 去重优先级：
+ * 1. PENDING/IN_PROGRESS 任务 → 不创建（任务处理中）
+ * 2. COMPLETED/FAILED 任务 → 不创建（已探索过）
+ * 3. 已存在的项目（通过 ExternalLink）→ 不创建（已收录）
+ * 4. 无任何记录 → 允许创建
+ *
+ * @param url - 要检查的 URL
+ * @returns 检查结果
+ */
+async function checkUrlDuplicate(url: string): Promise<{
+  url: string
+  shouldCreate: boolean
+  reason: string
+  existingTask?: {
+    id: string
+    status: string
+    sourceUrl: string
+    createdAt: Date
+    projectId?: string | null
+  }
+  existingProject?: {
+    id: string
+    name: string
+    slug: string
+  }
+}> {
+  // 优先级 1: 检查是否有 PENDING/IN_PROGRESS 的相同 URL 任务
+  const activeTask = await prisma.projectDiscoveryTask.findFirst({
+    where: {
+      sourceUrl: url,
+      status: {
+        in: ['PENDING', 'IN_PROGRESS'],
+      },
+    },
+    select: {
+      id: true,
+      status: true,
+      sourceUrl: true,
+      createdAt: true,
+      projectId: true,
+    },
+  })
+
+  if (activeTask) {
+    return {
+      url,
+      shouldCreate: false,
+      reason: `Task already exists with status ${activeTask.status}`,
+      existingTask: activeTask,
+    }
+  }
+
+  // 优先级 2: 检查是否有 COMPLETED/FAILED 的相同 URL 任务
+  const finishedTask = await prisma.projectDiscoveryTask.findFirst({
+    where: {
+      sourceUrl: url,
+      status: {
+        in: ['COMPLETED', 'FAILED'],
+      },
+    },
+    select: {
+      id: true,
+      status: true,
+      sourceUrl: true,
+      createdAt: true,
+      projectId: true,
+    },
+    orderBy: {
+      createdAt: 'desc',
+    },
+  })
+
+  if (finishedTask) {
+    // 如果任务已完成，返回项目信息
+    let projectInfo
+    if (finishedTask.status === 'COMPLETED' && finishedTask.projectId) {
+      const project = await prisma.project.findUnique({
+        where: { id: finishedTask.projectId },
+        select: {
+          id: true,
+          name: true,
+          slug: true,
+        },
+      })
+      if (project) {
+        projectInfo = project
+      }
+    }
+
+    return {
+      url,
+      shouldCreate: false,
+      reason:
+        finishedTask.status === 'COMPLETED'
+          ? 'Task already completed'
+          : 'Task already failed',
+      existingTask: finishedTask,
+      existingProject: projectInfo,
+    }
+  }
+
+  // 优先级 3: 检查 URL 对应的项目是否已存在（通过 ExternalLink）
+  const existingLink = await prisma.externalLink.findFirst({
+    where: {
+      url: url,
+    },
+    select: {
+      project: {
+        select: {
+          id: true,
+          name: true,
+          slug: true,
+        },
+      },
+    },
+  })
+
+  if (existingLink) {
+    return {
+      url,
+      shouldCreate: false,
+      reason: 'Project already exists with this URL',
+      existingProject: existingLink.project,
+    }
+  }
+
+  // 默认: 允许创建新任务
+  return {
+    url,
+    shouldCreate: true,
+    reason: 'No existing task or project found',
+  }
+}
+
+/**
+ * POST /api/discovery/check-duplicates
+ *
+ * 检查 URL 是否应该创建新的探索任务
+ *
+ * 请求体:
+ * {
+ *   "apiKey": "xxx",
+ *   "urls": ["https://github.com/user/repo", ...],
+ *   "sourceType": "manual"  // 可选
+ * }
+ *
+ * 返回:
+ * {
+ *   "success": true,
+ *   "results": [
+ *     {
+ *       "url": "...",
+ *       "shouldCreate": false,
+ *       "reason": "Task already exists with status PENDING",
+ *       "existingTask": {...},
+ *       "existingProject": {...}
+ *     }
+ *   ],
+ *   "stats": {
+ *     "total": 10,
+ *     "shouldCreate": 5,
+ *     "duplicate": 5
+ *   }
+ * }
+ */
+export async function POST(request: NextRequest) {
+  const startTime = Date.now()
+
+  try {
+    const body = await request.json()
+
+    // 验证请求体
+    const validationResult = CheckTaskDuplicatesSchema.safeParse(body)
+
+    if (!validationResult.success) {
+      return NextResponse.json(
+        {
+          success: false,
+          error: 'Validation error',
+          details: validationResult.error.errors.map((e) => e.message),
+        },
+        { status: 400 }
+      )
+    }
+
+    const { apiKey, urls, sourceType } = validationResult.data
+
+    // 验证 API Key
+    const validApiKey = process.env.WEBHOOK_API_KEY
+    if (
+      !validApiKey ||
+      !crypto.timingSafeEqual(
+        Buffer.from(apiKey),
+        Buffer.from(validApiKey)
+      )
+    ) {
+      return NextResponse.json(
+        {
+          success: false,
+          error: 'Unauthorized',
+          details: ['Invalid or missing API Key'],
+        },
+        { status: 401 }
+      )
+    }
+
+    // 并行检查所有 URL
+    const results = await Promise.all(urls.map((url) => checkUrlDuplicate(url)))
+
+    // 统计信息
+    const stats = {
+      total: results.length,
+      shouldCreate: results.filter((r) => r.shouldCreate).length,
+      duplicate: results.filter((r) => !r.shouldCreate).length,
+    }
+
+    const duration = Date.now() - startTime
+
+    console.warn(
+      `[CheckTaskDuplicates] Checked ${stats.total} URLs in ${duration}ms: ${stats.shouldCreate} should create, ${stats.duplicate} duplicate`
+    )
+
+    return NextResponse.json({
+      success: true,
+      results,
+      stats,
+    })
+  } catch (error) {
+    console.error('[CheckTaskDuplicates] Error:', error)
+    return NextResponse.json(
+      {
+        success: false,
+        error: 'Internal server error',
+        details: [error instanceof Error ? error.message : 'Unknown error'],
+      },
+      { status: 500 }
+    )
+  }
+}
diff --git a/src/lib/validations.ts b/src/lib/validations.ts
index 564e971..fcf5dbd 100644
--- a/src/lib/validations.ts
+++ b/src/lib/validations.ts
@@ -91,6 +91,15 @@ export const GetDiscoveryTasksQuerySchema = z.object({
   offset: z.coerce.number().int().nonnegative().default(0),
 })
 
+/**
+ * 检查任务去重 Schema
+ * 用于在创建任务前检查 URL 是否已存在
+ */
+export const CheckTaskDuplicatesSchema = WebhookAuthSchema.extend({
+  urls: z.array(z.string().url().max(2000)).min(1).max(100),
+  sourceType: z.string().max(50).optional(),
+})
+
 // ================================
 // Query Schemas
 // ================================