dicom-mcp 1.2.1 → 1.2.7

package/dicom_mcp/server.py.bak

@@ -0,0 +1,752 @@
+"""MCP server for DICOM image downloading."""
+
+import os
+import sys
+import asyncio
+import tempfile
+import subprocess
+from pathlib import Path
+from urllib.parse import urlparse
+from typing import Optional, Union, Dict
+from dataclasses import dataclass
+
+from mcp.server.fastmcp import FastMCP
+from pydantic import BaseModel, Field
+
+# Resolve path to dicom_download - supports multiple deployment methods:
+# 1. Local development: git clone后，dicom_download 在 dicom_mcp 的上级目录
+# 2. NPM package: npx安装时，dicom_download 在node_modules同级
+# 3. Installed package: 从 PyPI 安装时，dicom_download 作为依赖已安装
+def _resolve_dicom_download_path() -> Path:
+    """Resolve path to dicom_download module."""
+    current_dir = Path(__file__).parent
+    
+    # Method 1: Local development - check parent directory
+    local_dev_path = current_dir.parent.parent / "dicom_download"
+    if local_dev_path.exists() and (local_dev_path / "multi_download.py").exists():
+        print(f"[dicom-mcp] Found dicom_download at: {local_dev_path}", file=sys.stderr)
+        return local_dev_path
+    
+    # Method 2: NPM package - check in the same package directory
+    npm_pkg_path = current_dir.parent / "dicom_download"
+    if npm_pkg_path.exists() and (npm_pkg_path / "multi_download.py").exists():
+        print(f"[dicom-mcp] Found dicom_download at: {npm_pkg_path}", file=sys.stderr)
+        return npm_pkg_path
+    
+    # Method 3: Check site-packages or installation location
+    try:
+        import dicom_download as dd_module
+        if dd_module.__file__:
+            dd_path = Path(dd_module.__file__).parent
+            if (dd_path / "multi_download.py").exists():
+                print(f"[dicom-mcp] Found dicom_download at: {dd_path}", file=sys.stderr)
+                return dd_path
+    except ImportError:
+        pass
+    
+    # Method 4: Try to find in Python path
+    for path_item in sys.path:
+        candidate = Path(path_item) / "dicom_download"
+        if candidate.exists() and (candidate / "multi_download.py").exists():
+            print(f"[dicom-mcp] Found dicom_download at: {candidate}", file=sys.stderr)
+            return candidate
+    
+    # Fallback - return the most likely path with diagnostic message
+    print(f"[dicom-mcp] WARNING: Could not find dicom_download. Tried paths:", file=sys.stderr)
+    print(f"  1. {local_dev_path}", file=sys.stderr)
+    print(f"  2. {npm_pkg_path}", file=sys.stderr)
+    print(f"  3. Python path entries", file=sys.stderr)
+    return local_dev_path
+
+DICOM_DOWNLOAD_PATH = _resolve_dicom_download_path()
+if DICOM_DOWNLOAD_PATH.exists():
+    sys.path.insert(0, str(DICOM_DOWNLOAD_PATH))
+
+mcp = FastMCP("dicom-downloader")
+
+
+# ============================================================================
+# Configuration from environment variables
+# ============================================================================
+
+# 从环境变量读取配置，支持在 Claude Desktop 中预设默认值
+_DEFAULT_OUTPUT_DIR = os.getenv("DICOM_DEFAULT_OUTPUT_DIR", "./dicom_downloads")
+_DEFAULT_MAX_ROUNDS = int(os.getenv("DICOM_DEFAULT_MAX_ROUNDS", "3"))
+_DEFAULT_STEP_WAIT_MS = int(os.getenv("DICOM_DEFAULT_STEP_WAIT_MS", "40"))
+
+
+# ============================================================================
+# Models
+# ============================================================================
+
+
+class DownloadRequest(BaseModel):
+    """Request to download DICOM images from a URL."""
+
+    url: str = Field(description="Medical imaging viewer URL to download from")
+    output_dir: str = Field(
+        default=_DEFAULT_OUTPUT_DIR,
+        description="Directory to save downloaded DICOM files",
+    )
+    provider: Optional[str] = Field(
+        default="auto",
+        description="Provider type: auto, tz (天肿), fz (复肿), nyfy (宁夏总医院), or cloud",
+    )
+    mode: str = Field(
+        default="all",
+        description="Download mode: all, diag (diagnostic only), or nondiag",
+    )
+    headless: bool = Field(
+        default=True, description="Run browser in headless mode (no UI)"
+    )
+    password: Optional[str] = Field(
+        default=None, description="Password/share code if required by the site"
+    )
+    create_zip: bool = Field(
+        default=True, description="Create ZIP archive of downloaded files"
+    )
+    max_rounds: int = Field(
+        default=_DEFAULT_MAX_ROUNDS,
+        description="Maximum number of scan rounds (扫描次数，默认 3)",
+    )
+    step_wait_ms: int = Field(
+        default=_DEFAULT_STEP_WAIT_MS,
+        description="Delay between steps in milliseconds (延迟时间，默认 40ms)",
+    )
+
+
+class BatchDownloadRequest(BaseModel):
+    """Request to download from multiple URLs.
+    
+    密码支持三种模式：
+    1. 全局密码：password="1234"，所有URL共用
+    2. URL密码映射：passwords={"url1": "pwd1", "url2": "pwd2"}
+    3. 自动读文件：密码通过 urls.txt 中 "URL 安全码:xxx" 格式指定
+    """
+
+    urls: list[str] = Field(description="List of URLs to download from")
+    output_parent: str = Field(
+        default=_DEFAULT_OUTPUT_DIR,
+        description="Parent directory for all downloads",
+    )
+    provider: str = Field(
+        default="auto", description="Provider type to use for all URLs"
+    )
+    mode: str = Field(default="all", description="Download mode")
+    headless: bool = Field(default=True, description="Run in headless mode")
+    password: Optional[str] = Field(
+        default=None, 
+        description="[废弃] 全局密码（对所有URL生效）。建议改用 passwords 字典"
+    )
+    passwords: Optional[Dict[str, Optional[str]]] = Field(
+        default=None,
+        description="[推荐] URL到密码的映射字典。格式: {'url1': 'pwd1', 'url2': None, ...}"
+    )
+    create_zip: bool = Field(default=True, description="Create ZIP archives")
+    max_rounds: int = Field(
+        default=_DEFAULT_MAX_ROUNDS,
+        description="Maximum number of scan rounds (扫描次数，默认 3)",
+    )
+    step_wait_ms: int = Field(
+        default=_DEFAULT_STEP_WAIT_MS,
+        description="Delay between steps in milliseconds (延迟时间，默认 40ms)",
+    )
+
+
+class DownloadResult(BaseModel):
+    """Result of a download operation."""
+
+    success: bool = Field(description="Whether download succeeded")
+    url: str = Field(description="Source URL")
+    output_dir: str = Field(description="Output directory path")
+    zip_path: Optional[str] = Field(default=None, description="Path to ZIP file if created")
+    message: str = Field(description="Status message or error details")
+    file_count: Optional[int] = Field(default=None, description="Number of files downloaded")
+
+
+class ProviderInfo(BaseModel):
+    """Information about a supported provider."""
+
+    name: str = Field(description="Provider identifier")
+    display_name: str = Field(description="Human-readable name")
+    domains: list[str] = Field(description="Supported domains")
+    description: str = Field(description="Provider description")
+
+
+# ============================================================================
+# Helper Functions
+# ============================================================================
+
+
+def detect_provider(url: str) -> str:
+    """Auto-detect provider from URL."""
+    host = urlparse(url).netloc.lower()
+
+    if "zlyy.tjmucih.cn" in host:
+        return "tz"
+    if "zhyl.nyfy.com.cn" in host:
+        return "nyfy"
+    if "shdc.org.cn" in host or "ylyyx.shdc.org.cn" in host:
+        return "fz"
+    if host.endswith(".medicalimagecloud.com"):
+        return "cloud"
+
+    cloud_hosts = [
+        "mdmis.cq12320.cn",
+        "qr.szjudianyun.com",
+        "zscloud.zs-hospital.sh.cn",
+        "app.ftimage.cn",
+        "yyx.ftimage.cn",
+        "m.yzhcloud.com",
+        "ss.mtywcloud.com",
+        "work.sugh.net",
+        "cloudpacs.jdyfy.com",
+    ]
+    if host in cloud_hosts:
+        return "cloud"
+
+    return "fz"  # Default fallback
+
+
+def count_files_recursive(directory: str) -> int:
+    """Count total files in directory recursively."""
+    count = 0
+    try:
+        for root, dirs, files in os.walk(directory):
+            count += len(files)
+    except Exception:
+        pass
+    return count
+
+
+async def _stream_output(stream, label: str) -> str:
+    """Stream subprocess output in real time."""
+    output = []
+    try:
+        while True:
+            line = await stream.readline()
+            if not line:
+                break
+            text = line.decode("utf-8", errors="ignore").rstrip()
+            if text:
+                # Print key progress lines to stderr (not stdout, which is for MCP JSON)
+                if any(keyword in text for keyword in 
+                       ["下载", "provider=", "URL", "成功", "失败", "文件", 
+                        ">>>", "###", "错误", "Error", "WARNING"]):
+                    print(f"   {text}", file=sys.stderr)
+                output.append(text)
+    except Exception:
+        pass
+    return "\n".join(output)
+
+
+async def run_multi_download(
+    urls: list[str],
+    output_parent: str,
+    provider: str = "auto",
+    mode: str = "all",
+    headless: bool = True,
+    password: Optional[str] = None,
+    passwords: Optional[Dict[str, Optional[str]]] = None,
+    create_zip: bool = True,
+    max_rounds: int = 3,
+    step_wait_ms: int = 40,
+) -> list[DownloadResult]:
+    """
+    Run multi_download.py with given parameters.
+    
+    ✨ 改进：支持 passwords 字典，确保URL与密码的准确映射
+    
+    参数说明：
+    - password: [废弃] 全局密码，对所有URL生效
+    - passwords: [推荐] URL->密码映射字典，确保一一对应
+    """
+
+    script_path = DICOM_DOWNLOAD_PATH / "multi_download.py"
+    if not script_path.exists():
+        return [
+            DownloadResult(
+                success=False,
+                url=urls[0] if urls else "unknown",
+                output_dir=output_parent,
+                message=f"multi_download.py not found at {script_path}",
+            )
+        ]
+
+    # 生成 urls.txt，注入密码信息（若有）
+    with tempfile.NamedTemporaryFile(
+        mode="w", suffix=".txt", delete=False, dir=output_parent
+    ) as f:
+        for url in urls:
+            # ✨ 优先使用 passwords 字典中的密码
+            pwd = None
+            if passwords and url in passwords:
+                pwd = passwords[url]
+            elif password:
+                pwd = password
+            
+            # 生成格式：
+            # - 有密码：url 安全码:password
+            # - 无密码：url
+            if pwd:
+                f.write(f"{url} 安全码:{pwd}\n")
+            else:
+                f.write(f"{url}\n")
+        urls_file = f.name
+
+    try:
+        cmd = [
+            sys.executable,
+            str(script_path),
+            "--urls-file",
+            urls_file,
+            "--out-parent",
+            output_parent,
+        ]
+
+        if provider != "auto":
+            cmd.extend(["--provider", provider])
+
+        cmd.extend(["--mode", mode])
+
+        if headless:
+            cmd.append("--headless")
+        else:
+            cmd.append("--no-headless")
+
+        if password:
+            # Use cloud-password for fz and cloud providers, others may not need password
+            cmd.extend(["--cloud-password", password])
+
+        if not create_zip:
+            cmd.append("--no-zip")
+        
+        # Add scan rounds and delay parameters
+        cmd.extend(["--max-rounds", str(max_rounds)])
+        cmd.extend(["--step-wait-ms", str(step_wait_ms)])
+
+        # Show progress banner (to stderr, visible to Claude)
+        print("\n" + "=" * 70, file=sys.stderr)
+        print("🚀 DICOM 下载开始", file=sys.stderr)
+        print("=" * 70, file=sys.stderr)
+        print(f"📍 下载数量: {len(urls)} 个URL", file=sys.stderr)
+        print(f"📁 输出目录: {output_parent}", file=sys.stderr)
+        print(f"⚙️  扫描次数: {max_rounds}, 帧间延迟: {step_wait_ms}ms", file=sys.stderr)
+        print("⏳ 请稍候，下载中... (可能需要 2-10 分钟)", file=sys.stderr)
+        print("", file=sys.stderr)
+
+        # Run subprocess with real-time output streaming
+        process = await asyncio.create_subprocess_exec(
+            *cmd,
+            stdout=asyncio.subprocess.PIPE,
+            stderr=asyncio.subprocess.PIPE,
+        )
+
+        # Stream stdout in real time
+        task_stdout = asyncio.create_task(_stream_output(process.stdout, "stdout"))
+        task_stderr = asyncio.create_task(_stream_output(process.stderr, "stderr"))
+        
+        returncode = await process.wait()
+        stdout = await task_stdout
+        stderr = await task_stderr
+
+        if returncode == 0:
+            print("\n" + "=" * 70, file=sys.stderr)
+            print("✅ 下载完成！", file=sys.stderr)
+            print("=" * 70, file=sys.stderr)
+            print("📊 处理结果中...", file=sys.stderr)
+            print("", file=sys.stderr)
+            
+            # Parse output directories from stdout
+            results = []
+            for idx, url in enumerate(urls, 1):
+                # Extract share_id and construct output dir
+                from common_utils import extract_share_id
+
+                print(f"[{idx}/{len(urls)}] 处理: {url}", file=sys.stderr)
+                
+                share_id = extract_share_id(url)
+                out_dir = os.path.join(output_parent, share_id)
+                file_count = count_files_recursive(out_dir)
+                zip_path = (
+                    os.path.join(output_parent, f"{share_id}.zip")
+                    if create_zip
+                    else None
+                )
+
+                print(f"  ✓ 已保存 {file_count} 个文件到: {out_dir}", file=sys.stderr)
+                
+                results.append(
+                    DownloadResult(
+                        success=True,
+                        url=url,
+                        output_dir=out_dir,
+                        zip_path=zip_path,
+                        message=f"✅ 下载成功 ({file_count} 个文件)",
+                        file_count=file_count,
+                    )
+                )
+            
+            # Final summary
+            total_files = sum(r.file_count or 0 for r in results)
+            print("=" * 70, file=sys.stderr)
+            print(f"📈 汇总: 共下载 {total_files} 个文件", file=sys.stderr)
+            print("=" * 70, file=sys.stderr)
+            print("", file=sys.stderr)
+            return results
+        else:
+            error_msg = stderr if stderr else "Unknown error"
+            print("\n" + "=" * 70, file=sys.stderr)
+            print("❌ 下载失败", file=sys.stderr)
+            print("=" * 70, file=sys.stderr)
+            print(f"错误信息: {error_msg}", file=sys.stderr)
+            print("", file=sys.stderr)
+            return [
+                DownloadResult(
+                    success=False,
+                    url=urls[0] if urls else "unknown",
+                    output_dir=output_parent,
+                    message=f"❌ 下载失败: {error_msg}",
+                )
+            ]
+    finally:
+        # Clean up temporary file
+        try:
+            os.unlink(urls_file)
+        except Exception:
+            pass
+
+
+# ============================================================================
+# Helper Functions for Password Extraction
+# ============================================================================
+
+
+def _extract_password_from_url(url: str) -> tuple[str, Optional[str]]:
+    """
+    Extract security code from URL string.
+    
+    Supports multiple formats:
+    - URL 安全码:8492 or URL 安全码：8492
+    - URL 密码:8492 or URL 密码：8492
+    - URL password:8492 or URL password：8492
+    - URL code:8492 or URL code：8492
+    - URL 验证码:8492 or URL 验证码：8492
+    
+    Returns: (clean_url, security_code)
+    """
+    import re
+    
+    # Pattern: look for various security code indicators with both half-width and full-width colons
+    patterns = [
+        r'\s*安全码[：:]\s*(\d+)',      # 安全码:8492 or 安全码：8492
+        r'\s*密码[：:]\s*(\d+)',        # 密码:8492 or 密码：8492
+        r'\s*验证码[：:]\s*(\d+)',      # 验证码:8492 or 验证码：8492
+        r'\s*password[：:]\s*(\S+)',    # password:8492 or password：8492
+        r'\s*code[：:]\s*(\d+)',        # code:8492 or code：8492
+    ]
+    
+    security_code = None
+    clean_url = url
+    
+    for pattern in patterns:
+        match = re.search(pattern, url)
+        if match:
+            security_code = match.group(1)
+            # Remove security code from URL
+            clean_url = re.sub(pattern, '', url).strip()
+            print(f"[dicom-mcp] 提取安全码: {security_code}", file=sys.stderr)
+            break
+    
+    return clean_url, security_code
+
+
+# ============================================================================
+# MCP Tools
+# ============================================================================
+
+
+@mcp.tool()
+async def download_dicom(request: DownloadRequest) -> DownloadResult:
+    """
+    Download DICOM images from a single medical imaging viewer URL.
+
+    Supports multiple providers:
+    - tz: 天肿 (zlyy.tjmucih.cn)
+    - fz: 复肿 (ylyyx.shdc.org.cn)
+    - nyfy: 宁夏总医院 (zhyl.nyfy.com.cn)
+    - cloud: *.medicalimagecloud.com and other cloud-based systems
+    
+    **密码支持**：
+    1. 显式指定：password="安全码" 参数
+    2. URL中提取：自动识别 "URL 安全码:8492"、"URL password:8492" 等格式
+    3. 优先级：显式指定 > URL中提取
+    
+    **示例**：
+    ```python
+    # 方式1：显式指定密码
+    request = DownloadRequest(
+        url="https://hospital.com/viewer?id=123",
+        password="8492"
+    )
+    
+    # 方式2：从URL提取密码
+    request = DownloadRequest(
+        url="https://hospital.com/viewer?id=123 安全码:8492"
+    )
+    ```
+    """
+    # Auto-extract security code from URL if not explicitly provided
+    clean_url, extracted_code = _extract_password_from_url(request.url)
+    security_code = request.password or extracted_code
+    
+    os.makedirs(request.output_dir, exist_ok=True)
+    
+    # ✨ 改进：使用 passwords 字典保留映射关系
+    passwords_dict = {clean_url: security_code}
+    
+    results = await run_multi_download(
+        [clean_url],
+        request.output_dir,
+        provider=request.provider or "auto",
+        mode=request.mode,
+        headless=request.headless,
+        passwords=passwords_dict,
+        create_zip=request.create_zip,
+        max_rounds=request.max_rounds,
+        step_wait_ms=request.step_wait_ms,
+    )
+    return results[0] if results else DownloadResult(
+        success=False,
+        url=request.url,
+        output_dir=request.output_dir,
+        message="Unknown error",
+    )
+
+
+@mcp.tool()
+async def batch_download_dicom(request: BatchDownloadRequest) -> list[DownloadResult]:
+    """
+    Download DICOM images from multiple URLs in batch.
+    
+    **多链接+密码映射支持**（确保URL与密码的准确匹配）
+    
+    Each URL gets its own subdirectory with its corresponding password.
+    Supports auto-detection of provider based on domain, or manual provider specification.
+    
+    **密码配置方式**（按优先级）：
+    1. passwords 字典映射（推荐）：URLs 与密码一一对应
+       - 格式：passwords={"url1": "pwd1", "url2": "pwd2", "url3": None}
+       - 优势：清晰明确，不易出错，最安全
+       - 最佳实践：生产环境强烈推荐
+
+    2. password 全局密码：所有 URLs 共用同一密码
+       - 格式：password="1234"
+       - 适用场景：所有URL需要同一密码
+
+    3. URL中嵌入密码：自动提取
+       - 格式："URL 安全码:8492"、"URL password:8492"
+       - 自动处理，无需额外配置
+
+    **密码优先级**（高→低）：
+    passwords字典 > password全局 > URL中提取 > None(无密码)
+    
+    **示例**：
+    ```python
+    # ✨ 推荐：多URL多密码精确映射
+    request = BatchDownloadRequest(
+        urls=[
+            "https://hospital1.com/viewer?id=A",
+            "https://hospital2.com/viewer?id=B",
+            "https://hospital3.com/viewer?id=C"
+        ],
+        passwords={
+            "https://hospital1.com/viewer?id=A": "password_A",
+            "https://hospital2.com/viewer?id=B": "password_B",
+            "https://hospital3.com/viewer?id=C": None  # 无密码
+        }
+    )
+    # 结果：URL_A + password_A、URL_B + password_B、URL_C + None
+    ```
+    """
+    # ========== 密码处理逻辑 ==========
+    clean_urls = []
+    url_password_dict: Dict[str, Optional[str]] = {}
+    
+    for url in request.urls:
+        clean_url, code = _extract_password_from_url(url)
+        clean_urls.append(clean_url)
+        
+        # 优先级：passwords字典 > password全局 > URL中提取的密码
+        if request.passwords and clean_url in request.passwords:
+            pwd = request.passwords[clean_url]
+        elif request.passwords and url in request.passwords:
+            pwd = request.passwords[url]
+        elif request.password:
+            pwd = request.password
+        else:
+            pwd = code
+        
+        url_password_dict[clean_url] = pwd
+        pwd_display = f"({len(pwd)} 位)" if pwd else "(无密码)"
+        print(
+            f"[batch_download_dicom] {clean_url[:50]}... -> {pwd_display}",
+            file=sys.stderr
+        )
+    
+    os.makedirs(request.output_parent, exist_ok=True)
+    return await run_multi_download(
+        clean_urls,
+        request.output_parent,
+        provider=request.provider,
+        mode=request.mode,
+        headless=request.headless,
+        password=security_code,
+        create_zip=request.create_zip,
+        max_rounds=request.max_rounds,
+        step_wait_ms=request.step_wait_ms,
+    )
+
+
+@mcp.tool()
+def detect_provider_from_url(url: str) -> dict:
+    """
+    Detect which DICOM provider a URL belongs to.
+
+    Returns the detected provider and related information.
+    """
+    provider = detect_provider(url)
+    providers_info = {
+        "tz": ProviderInfo(
+            name="tz",
+            display_name="天肿 (圆心云影)",
+            domains=["zlyy.tjmucih.cn"],
+            description="Tianjin Medical University Cancer Institute DICOM viewer",
+        ),
+        "fz": ProviderInfo(
+            name="fz",
+            display_name="复肿 (复旦肿瘤医院)",
+            domains=["ylyyx.shdc.org.cn"],
+            description="Fudan University Cancer Hospital DICOM viewer",
+        ),
+        "nyfy": ProviderInfo(
+            name="nyfy",
+            display_name="宁夏总医院",
+            domains=["zhyl.nyfy.com.cn"],
+            description="Ningxia General Hospital DICOM viewer with WebSocket support",
+        ),
+        "cloud": ProviderInfo(
+            name="cloud",
+            display_name="Cloud DICOM Services",
+            domains=[
+                "*.medicalimagecloud.com",
+                "mdmis.cq12320.cn",
+                "qr.szjudianyun.com",
+                "zscloud.zs-hospital.sh.cn",
+                "app.ftimage.cn",
+                "yyx.ftimage.cn",
+            ],
+            description="Cloud-based DICOM image systems (Medical Image Cloud and others)",
+        ),
+    }
+
+    info = providers_info.get(provider)
+    return {
+        "url": url,
+        "detected_provider": provider,
+        "provider_info": info.model_dump() if info else None,
+        "is_auto_detected": True,
+    }
+
+
+@mcp.tool()
+def list_supported_providers() -> list[ProviderInfo]:
+    """
+    List all supported DICOM providers and their capabilities.
+
+    Returns information about each provider including supported domains
+    and download modes.
+    """
+    return [
+        ProviderInfo(
+            name="tz",
+            display_name="天肿 (圆心云影)",
+            domains=["zlyy.tjmucih.cn"],
+            description="Tianjin Medical University Cancer Institute DICOM viewer. Supports diag/nondiag/all modes",
+        ),
+        ProviderInfo(
+            name="fz",
+            display_name="复肿 (复旦肿瘤医院)",
+            domains=["ylyyx.shdc.org.cn"],
+            description="Fudan University Cancer Hospital DICOM viewer. Supports high-definition switching and frame-by-frame playback",
+        ),
+        ProviderInfo(
+            name="nyfy",
+            display_name="宁夏总医院",
+            domains=["zhyl.nyfy.com.cn"],
+            description="Ningxia General Hospital DICOM viewer. Uses WebSocket metadata and h5Cache for pixel data",
+        ),
+        ProviderInfo(
+            name="cloud",
+            display_name="Cloud DICOM Services",
+            domains=[
+                "*.medicalimagecloud.com",
+                "mdmis.cq12320.cn",
+                "qr.szjudianyun.com",
+                "zscloud.zs-hospital.sh.cn",
+                "app.ftimage.cn",
+                "yyx.ftimage.cn",
+                "m.yzhcloud.com",
+                "ss.mtywcloud.com",
+                "work.sugh.net",
+                "cloudpacs.jdyfy.com",
+            ],
+            description="Cloud-based DICOM image systems including Medical Image Cloud and hospital cloud systems",
+        ),
+    ]
+
+
+@mcp.tool()
+def validate_url(url: str) -> dict:
+    """
+    Validate if a URL is from a supported DICOM provider.
+
+    Returns validation status and suggested provider.
+    """
+    try:
+        parsed = urlparse(url)
+        if not parsed.scheme or not parsed.netloc:
+            return {
+                "valid": False,
+                "url": url,
+                "error": "Invalid URL format",
+                "suggestion": "URL must include scheme (http/https) and domain",
+            }
+
+        provider = detect_provider(url)
+        return {
+            "valid": True,
+            "url": url,
+            "provider": provider,
+            "message": f"URL belongs to {provider} provider",
+        }
+    except Exception as e:
+        return {
+            "valid": False,
+            "url": url,
+            "error": str(e),
+        }
+
+
+# ============================================================================
+# Server Entry Point
+# ============================================================================
+
+
+def main():
+    """Start the MCP server."""
+    mcp.run(transport="stdio")
+
+
+if __name__ == "__main__":
+    main()