redfox-mcp 0.1.0__tar.gz

redfox_mcp-0.1.0/.gitignore

@@ -0,0 +1,6 @@
+__pycache__/
+*.pyc
+.venv/
+dist/
+*.egg-info/
+.DS_Store

redfox_mcp-0.1.0/PKG-INFO

@@ -0,0 +1,113 @@
+Metadata-Version: 2.5
+Name: redfox-mcp
+Version: 0.1.0
+Summary: RedFoxHub MCP server: Douyin/Xiaohongshu/WeChat/Bilibili/Toutiao/TikTok data APIs + AI search, image and video generation tools
+Project-URL: Homepage, https://github.com/redfox-data/redfox-mcp
+Project-URL: Website, https://redfox.hk/apis/?source=mcp
+License-Expression: MIT
+Keywords: bilibili,douyin,mcp,redfox,tiktok,toutiao,wechat,xiaohongshu
+Requires-Python: >=3.10
+Requires-Dist: cryptography<49; platform_machine == 'x86_64' and sys_platform == 'darwin'
+Requires-Dist: fastmcp>=2.0
+Requires-Dist: httpx>=0.24
+Requires-Dist: redfox-python-sdk
+Description-Content-Type: text/markdown
+
+<p align="center">
+  <a href="https://redfox.hk/?source=github"><img src="https://lyy.redfox.hk/page/logo-redfox-name.png" alt="RedFox Logo" width="200"></a>
+</p>
+
+<p align="right">
+  <a href="https://github.com/redfox-data/redfox-mcp/blob/main/README.zh.md">中文</a>
+  English
+</p>
+
+# redfox-mcp
+
+RedFoxHub MCP Server — turns the data APIs of 6 major content platforms plus AI search / generation capabilities into 40 MCP tools, ready for any MCP client such as dsh, Claude Code or Cursor.
+
+## Tools (40)
+
+| Category | Tools | Notes |
+|---|---|---|
+| Douyin | `douyin_search_articles` / `douyin_search_users` / `douyin_get_user` / `douyin_get_user_works` / `douyin_get_work` / `douyin_search_ai_articles` | work search, account search, account info, work lists, work detail, AI-work feed |
+| Xiaohongshu | `xiaohongshu_search_articles` / `xiaohongshu_search_users` / `xiaohongshu_get_account` / `xiaohongshu_get_work` / `xiaohongshu_search_ai_articles` | note search, creator search, account info, note detail, AI-note feed |
+| WeChat Official Accounts | `wechat_search_articles` / `wechat_search_users` / `wechat_get_account` / `wechat_get_user_works` / `wechat_get_work` / `wechat_get_article_detail` / `wechat_search_ai_articles` | article search (incl. full-text detail), account search, article lists, AI-article feed |
+| Bilibili | `bilibili_search_articles` / `bilibili_search_users` / `bilibili_get_account` / `bilibili_get_user_works` / `bilibili_get_work` | video search, UP-master search / info / videos, video detail |
+| Toutiao | `toutiao_search_works` / `toutiao_get_work` | content search, work detail (realtime) |
+| TikTok | `tiktok_search_users` | account search |
+| AI search | `ai_search_kimi` / `ai_search_doubao` / `ai_search_deepseek` | one call submits the query and waits for the full answer |
+| AI generation | `gpt_image_generate` / `doubao_image_pro_generate` / `doubao_image_lite_generate` / `doubao_video_generate` | text-to-image / image-to-image / image sets / text-to-video, one call submits and waits |
+| Task follow-up | `ai_search_*_result` / `gpt_image_result` / `doubao_image_*_result` / `doubao_video_result` (7 in total) | when an async tool times out it returns a `taskId`; use the matching result tool to fetch the outcome |
+
+Async tools (AI search / generation) poll internally: submit → wait → return the full result, with no manual `taskId` handling. If the wait exceeds `timeout_seconds` (default 240s, 480s for video), a `taskId` is returned for the matching result tool.
+
+## Authentication
+
+All APIs require a RedFoxHub API key:
+
+1. Get one at <https://redfox.hk/settings/api-keys>
+2. Set the environment variable:
+
+```bash
+export REDFOX_API_KEY="YOUR_API_KEY"
+```
+
+Without a key, every tool returns a structured message explaining how to obtain one.
+
+## Install & Run
+
+Python ≥ 3.10 required. Recommended via [uv](https://docs.astral.sh/uv/):
+
+```bash
+uvx redfox-mcp
+```
+
+or:
+
+```bash
+pip install redfox-mcp
+redfox-mcp
+```
+
+The server runs on stdio transport.
+
+## Client Configuration
+
+### dsh (DeepSeek Harness)
+
+Install the official bundle plugin [redfox-community-dsh](https://github.com/redfox-data/redfox-community-dsh) — it registers this MCP server out of the box, exposing tools as `mcp__redfox__*`:
+
+```bash
+dsh plugin --profile web add -w github:redfox-data/redfox-community-dsh
+```
+
+### Claude Code
+
+```bash
+claude mcp add redfox --env REDFOX_API_KEY=YOUR_API_KEY -- uvx redfox-mcp
+```
+
+### Cursor / other MCP clients
+
+Add to your MCP configuration:
+
+```json
+{
+  "mcpServers": {
+    "redfox": {
+      "command": "uvx",
+      "args": ["redfox-mcp"],
+      "env": { "REDFOX_API_KEY": "YOUR_API_KEY" }
+    }
+  }
+}
+```
+
+## Under the Hood
+
+Built on the official SDK [redfox-python-sdk](https://github.com/redfox-data/redfox-python-sdk). API docs: <https://redfox.hk>.
+
+## License
+
+MIT

redfox_mcp-0.1.0/README.md

@@ -0,0 +1,98 @@
+<p align="center">
+  <a href="https://redfox.hk/?source=github"><img src="https://lyy.redfox.hk/page/logo-redfox-name.png" alt="RedFox Logo" width="200"></a>
+</p>
+
+<p align="right">
+  <a href="https://github.com/redfox-data/redfox-mcp/blob/main/README.zh.md">中文</a>
+  English
+</p>
+
+# redfox-mcp
+
+RedFoxHub MCP Server — turns the data APIs of 6 major content platforms plus AI search / generation capabilities into 40 MCP tools, ready for any MCP client such as dsh, Claude Code or Cursor.
+
+## Tools (40)
+
+| Category | Tools | Notes |
+|---|---|---|
+| Douyin | `douyin_search_articles` / `douyin_search_users` / `douyin_get_user` / `douyin_get_user_works` / `douyin_get_work` / `douyin_search_ai_articles` | work search, account search, account info, work lists, work detail, AI-work feed |
+| Xiaohongshu | `xiaohongshu_search_articles` / `xiaohongshu_search_users` / `xiaohongshu_get_account` / `xiaohongshu_get_work` / `xiaohongshu_search_ai_articles` | note search, creator search, account info, note detail, AI-note feed |
+| WeChat Official Accounts | `wechat_search_articles` / `wechat_search_users` / `wechat_get_account` / `wechat_get_user_works` / `wechat_get_work` / `wechat_get_article_detail` / `wechat_search_ai_articles` | article search (incl. full-text detail), account search, article lists, AI-article feed |
+| Bilibili | `bilibili_search_articles` / `bilibili_search_users` / `bilibili_get_account` / `bilibili_get_user_works` / `bilibili_get_work` | video search, UP-master search / info / videos, video detail |
+| Toutiao | `toutiao_search_works` / `toutiao_get_work` | content search, work detail (realtime) |
+| TikTok | `tiktok_search_users` | account search |
+| AI search | `ai_search_kimi` / `ai_search_doubao` / `ai_search_deepseek` | one call submits the query and waits for the full answer |
+| AI generation | `gpt_image_generate` / `doubao_image_pro_generate` / `doubao_image_lite_generate` / `doubao_video_generate` | text-to-image / image-to-image / image sets / text-to-video, one call submits and waits |
+| Task follow-up | `ai_search_*_result` / `gpt_image_result` / `doubao_image_*_result` / `doubao_video_result` (7 in total) | when an async tool times out it returns a `taskId`; use the matching result tool to fetch the outcome |
+
+Async tools (AI search / generation) poll internally: submit → wait → return the full result, with no manual `taskId` handling. If the wait exceeds `timeout_seconds` (default 240s, 480s for video), a `taskId` is returned for the matching result tool.
+
+## Authentication
+
+All APIs require a RedFoxHub API key:
+
+1. Get one at <https://redfox.hk/settings/api-keys>
+2. Set the environment variable:
+
+```bash
+export REDFOX_API_KEY="YOUR_API_KEY"
+```
+
+Without a key, every tool returns a structured message explaining how to obtain one.
+
+## Install & Run
+
+Python ≥ 3.10 required. Recommended via [uv](https://docs.astral.sh/uv/):
+
+```bash
+uvx redfox-mcp
+```
+
+or:
+
+```bash
+pip install redfox-mcp
+redfox-mcp
+```
+
+The server runs on stdio transport.
+
+## Client Configuration
+
+### dsh (DeepSeek Harness)
+
+Install the official bundle plugin [redfox-community-dsh](https://github.com/redfox-data/redfox-community-dsh) — it registers this MCP server out of the box, exposing tools as `mcp__redfox__*`:
+
+```bash
+dsh plugin --profile web add -w github:redfox-data/redfox-community-dsh
+```
+
+### Claude Code
+
+```bash
+claude mcp add redfox --env REDFOX_API_KEY=YOUR_API_KEY -- uvx redfox-mcp
+```
+
+### Cursor / other MCP clients
+
+Add to your MCP configuration:
+
+```json
+{
+  "mcpServers": {
+    "redfox": {
+      "command": "uvx",
+      "args": ["redfox-mcp"],
+      "env": { "REDFOX_API_KEY": "YOUR_API_KEY" }
+    }
+  }
+}
+```
+
+## Under the Hood
+
+Built on the official SDK [redfox-python-sdk](https://github.com/redfox-data/redfox-python-sdk). API docs: <https://redfox.hk>.
+
+## License
+
+MIT

redfox_mcp-0.1.0/pyproject.toml

@@ -0,0 +1,32 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "redfox-mcp"
+version = "0.1.0"
+description = "RedFoxHub MCP server: Douyin/Xiaohongshu/WeChat/Bilibili/Toutiao/TikTok data APIs + AI search, image and video generation tools"
+readme = "README.md"
+requires-python = ">=3.10"
+license = "MIT"
+keywords = ["mcp", "redfox", "douyin", "xiaohongshu", "wechat", "bilibili", "toutiao", "tiktok"]
+dependencies = [
+    "fastmcp>=2.0",
+    # Intel Mac 上 cryptography 49+ 无预编译 wheel 会触发源码构建，钉住 48.x
+    "cryptography<49; platform_machine == 'x86_64' and sys_platform == 'darwin'",
+    "redfox-python-sdk",
+    "httpx>=0.24",
+]
+
+[project.urls]
+Homepage = "https://github.com/redfox-data/redfox-mcp"
+Website = "https://redfox.hk/apis/?source=mcp"
+
+[project.scripts]
+redfox-mcp = "redfox_mcp.server:main"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/redfox_mcp"]
+
+[tool.hatch.build.targets.sdist]
+include = ["src/redfox_mcp", "README.md"]

redfox_mcp-0.1.0/src/redfox_mcp/__init__.py

@@ -0,0 +1,3 @@
+"""RedFox MCP Server — 红狐数据平台 MCP 工具服务"""
+
+__version__ = "0.1.0"

redfox_mcp-0.1.0/src/redfox_mcp/__main__.py

@@ -0,0 +1,6 @@
+"""支持 python -m redfox_mcp 启动"""
+
+from redfox_mcp.server import main
+
+if __name__ == "__main__":
+    main()

redfox_mcp-0.1.0/src/redfox_mcp/server.py

@@ -0,0 +1,520 @@
+"""RedFox MCP Server
+
+将 RedFoxHub（红狐数据平台）的 40 个 API 能力暴露为 MCP 工具：
+- 6 大内容平台数据查询（抖音/小红书/公众号/B站/头条/TikTok，26 个同步工具）
+- AI 搜索（Kimi/豆包/Deepseek，3 个自动轮询工具 + 3 个结果查询工具）
+- AI 生成（GPT 图片/豆包图片 Pro/Lite/豆包视频，4 个自动轮询工具 + 4 个结果查询工具）
+
+认证：环境变量 REDFOX_API_KEY（获取地址 https://redfox.hk/settings/api-keys?source=mcp）
+"""
+
+import time
+from typing import Any, Dict, List, Optional
+
+from fastmcp import FastMCP
+from redfox import RedFoxClient
+from redfox.exceptions import RedFoxAPIError, RedFoxAuthError, RedFoxRateLimitError
+
+mcp = FastMCP("redfox")
+
+API_KEY_GUIDE = (
+    "REDFOX_API_KEY 未配置或无效。请前往 "
+    "https://redfox.hk/settings/api-keys?source=mcp 注册并获取 API Key，"
+    "然后设置环境变量 REDFOX_API_KEY 后重启本服务。"
+)
+
+TASK_PENDING_MSG = (
+    "任务仍在进行中，已超过本次等待时间。"
+    "请稍后使用对应的 result 工具并传入此 taskId 查询结果。"
+)
+
+TERMINAL_STATUSES = {
+    "succeeded", "success", "completed", "complete", "done",
+    "failed", "error", "cancelled", "canceled",
+}
+
+_client: Optional[RedFoxClient] = None
+
+
+def _get_client() -> RedFoxClient:
+    global _client
+    if _client is None:
+        _client = RedFoxClient()  # 零配置：自动读环境变量 REDFOX_API_KEY
+    return _client
+
+
+def _call(fn_factory, **kwargs) -> Dict[str, Any]:
+    """统一调用 SDK 方法并把异常转为结构化结果，agent 可直接读取出错引导"""
+    try:
+        return fn_factory()(**{k: v for k, v in kwargs.items() if v is not None})
+    except (RedFoxAuthError, ValueError):
+        return {"error": "auth_failed", "message": API_KEY_GUIDE}
+    except RedFoxRateLimitError:
+        return {"error": "rate_limited", "message": "请求频率超限，请稍后重试"}
+    except RedFoxAPIError as e:
+        return {"error": "api_error", "code": e.code, "message": e.message}
+
+
+def _is_done(res: Any) -> bool:
+    """判断异步任务是否到达终态（兼容各品类不同的响应结构）"""
+    if not isinstance(res, dict):
+        return True
+    if res.get("error"):
+        return True
+    if res.get("completed") is True:
+        return True
+    status = res.get("status")
+    if isinstance(status, str) and status.lower() in TERMINAL_STATUSES:
+        return True
+    for field in ("content", "imagePaths", "images", "videoUrl", "videoUrls", "video"):
+        if res.get(field):
+            return True
+    return False
+
+
+def _poll(result_fn, task_id: str, timeout_seconds: int, interval: float = 3.0) -> Dict[str, Any]:
+    deadline = time.monotonic() + timeout_seconds
+    while True:
+        res = _call(result_fn, task_id=task_id)
+        if _is_done(res):
+            return res
+        if time.monotonic() >= deadline:
+            return {"completed": False, "taskId": task_id, "message": TASK_PENDING_MSG}
+        time.sleep(interval)
+
+
+def _run_task(submit_fn, result_fn, timeout_seconds: int, **kwargs) -> Dict[str, Any]:
+    """提交异步任务并自动轮询至完成；超时则返回 taskId 供 result 工具后续查询"""
+    submitted = _call(submit_fn, **kwargs)
+    if not isinstance(submitted, dict) or submitted.get("error"):
+        return submitted
+    task_id = submitted.get("taskId") or submitted.get("task_id")
+    if not task_id:
+        return submitted
+    return _poll(result_fn, task_id, timeout_seconds)
+
+
+# ─── 抖音 ────────────────────────────────────────────────
+
+
+@mcp.tool()
+def douyin_search_articles(keyword: str, offset: int = 0,
+                           sort_type: Optional[str] = None) -> Dict[str, Any]:
+    """搜索抖音作品（优质库）。keyword 必填；offset 分页偏移从 0 开始、每次 +20；
+    sort_type 排序方式，如 "default"。返回含 total/hasMore/list。"""
+    return _call(lambda: _get_client().douyin.search_articles,
+                 keyword=keyword, offset=offset, sort_type=sort_type)
+
+
+@mcp.tool()
+def douyin_search_users(keyword: str, offset: int = 0,
+                        sort_type: Optional[str] = None) -> Dict[str, Any]:
+    """搜索抖音账号（优质库）。keyword 必填；offset 分页偏移从 0 开始。"""
+    return _call(lambda: _get_client().douyin.search_users,
+                 keyword=keyword, offset=offset, sort_type=sort_type)
+
+
+@mcp.tool()
+def douyin_get_user(account_id: str) -> Dict[str, Any]:
+    """获取抖音账号信息（优质库）。account_id 为抖音账号 ID，
+    支持 unique_id、short_id、uid 任一匹配。"""
+    return _call(lambda: _get_client().douyin.get_user, account_id=account_id)
+
+
+@mcp.tool()
+def douyin_get_user_works(account_id: Optional[str] = None,
+                          author_url: Optional[str] = None,
+                          sec_user_id: Optional[str] = None,
+                          offset: int = 0,
+                          sort_type: Optional[str] = None) -> Dict[str, Any]:
+    """获取抖音账号作品列表（优质库）。account_id（抖音号）/ author_url（主页链接）/
+    sec_user_id 至少传一个；offset 每页 +20；sort_type：0=默认，2=最新，4=最热。"""
+    return _call(lambda: _get_client().douyin.get_user_works,
+                 account_id=account_id, author_url=author_url,
+                 sec_user_id=sec_user_id, offset=offset, sort_type=sort_type)
+
+
+@mcp.tool()
+def douyin_get_work(work_id: Optional[str] = None,
+                    work_url: Optional[str] = None) -> Dict[str, Any]:
+    """获取抖音作品详情（优质库）。work_id 与 work_url（作品链接）至少传一个，
+    返回互动数据、作者信息等。"""
+    return _call(lambda: _get_client().douyin.get_work, work_id=work_id, work_url=work_url)
+
+
+@mcp.tool()
+def douyin_search_ai_articles(keyword: str, page_num: int = 1, page_size: int = 20,
+                              start_time: Optional[str] = None,
+                              end_time: Optional[str] = None) -> Dict[str, Any]:
+    """搜索抖音 AI 相关作品（优质库）。keyword 必填；
+    start_time/end_time 格式如 "2026-06-01 00:00:00"。"""
+    return _call(lambda: _get_client().douyin.search_ai_articles,
+                 keyword=keyword, page_num=page_num, page_size=page_size,
+                 start_time=start_time, end_time=end_time)
+
+
+# ─── 小红书 ──────────────────────────────────────────────
+
+
+@mcp.tool()
+def xiaohongshu_search_articles(keyword: str, offset: int = 0,
+                                sort_type: Optional[str] = None) -> Dict[str, Any]:
+    """搜索小红书笔记（优质库）。keyword 必填；offset 分页偏移。"""
+    return _call(lambda: _get_client().xiaohongshu.search_articles,
+                 keyword=keyword, offset=offset, sort_type=sort_type)
+
+
+@mcp.tool()
+def xiaohongshu_search_users(keyword: str, offset: int = 0,
+                             sort_type: Optional[str] = None) -> Dict[str, Any]:
+    """搜索小红书博主账号（优质库）。keyword 必填；offset 分页偏移。"""
+    return _call(lambda: _get_client().xiaohongshu.search_users,
+                 keyword=keyword, offset=offset, sort_type=sort_type)
+
+
+@mcp.tool()
+def xiaohongshu_get_account(account_id: str,
+                            user_id: Optional[str] = None) -> Dict[str, Any]:
+    """获取小红书账号信息（优质库）。account_id 为小红书号（必填），user_id 可选。"""
+    return _call(lambda: _get_client().xiaohongshu.get_account,
+                 account_id=account_id, user_id=user_id)
+
+
+@mcp.tool()
+def xiaohongshu_get_work(work_id: Optional[str] = None,
+                         work_link: Optional[str] = None) -> Dict[str, Any]:
+    """获取小红书笔记详情（优质库）。work_id 与 work_link（笔记链接）至少传一个。"""
+    return _call(lambda: _get_client().xiaohongshu.get_work,
+                 work_id=work_id, work_link=work_link)
+
+
+@mcp.tool()
+def xiaohongshu_search_ai_articles(keyword: str, page_num: int = 1,
+                                   page_size: int = 20,
+                                   start_time: Optional[str] = None,
+                                   end_time: Optional[str] = None,
+                                   source: Optional[str] = None) -> Dict[str, Any]:
+    """搜索小红书 AI 创作相关笔记（优质库）。keyword 必填；
+    start_time/end_time 格式如 "2026-06-01 00:00:00"；source 为来源平台（可选）。"""
+    return _call(lambda: _get_client().xiaohongshu.search_ai_articles,
+                 keyword=keyword, page_num=page_num, page_size=page_size,
+                 start_time=start_time, end_time=end_time, source=source)
+
+
+# ─── 公众号 ──────────────────────────────────────────────
+
+
+@mcp.tool()
+def wechat_search_articles(keyword: str, offset: int = 0,
+                           sort_type: Optional[str] = None) -> Dict[str, Any]:
+    """搜索公众号文章（优质库）。keyword 必填；offset 从 0 开始、每页 +20。"""
+    return _call(lambda: _get_client().wechat.search_articles,
+                 keyword=keyword, offset=offset, sort_type=sort_type)
+
+
+@mcp.tool()
+def wechat_search_users(keyword: str, offset: int = 0,
+                        sort_type: Optional[str] = None) -> Dict[str, Any]:
+    """搜索公众号账号（优质库）。keyword 必填；sort_type：_0=默认，_2=最新，_4=最热。"""
+    return _call(lambda: _get_client().wechat.search_users,
+                 keyword=keyword, offset=offset, sort_type=sort_type)
+
+
+@mcp.tool()
+def wechat_get_account(account: str,
+                       account_name: Optional[str] = None) -> Dict[str, Any]:
+    """获取公众号账号信息（优质库）。account 为公众号微信号（必填），
+    account_name 为公众号名称（可选）。"""
+    return _call(lambda: _get_client().wechat.get_account,
+                 account=account, account_name=account_name)
+
+
+@mcp.tool()
+def wechat_get_user_works(account: str, account_name: Optional[str] = None,
+                          offset: int = 0, sort_type: Optional[str] = None,
+                          publish_time_start: Optional[str] = None,
+                          publish_time_end: Optional[str] = None) -> Dict[str, Any]:
+    """获取公众号文章列表（优质库）。account 为公众号微信号（必填）；
+    sort_type：_0=默认，_2=最新，_4=最热；publish_time_start/end 格式如 "2026-07-01"。"""
+    return _call(lambda: _get_client().wechat.get_user_works,
+                 account=account, account_name=account_name, offset=offset,
+                 sort_type=sort_type, publish_time_start=publish_time_start,
+                 publish_time_end=publish_time_end)
+
+
+@mcp.tool()
+def wechat_get_work(work_uuid: str) -> Dict[str, Any]:
+    """根据作品 UUID 获取公众号文章元数据（优质库）。"""
+    return _call(lambda: _get_client().wechat.get_work, work_uuid=work_uuid)
+
+
+@mcp.tool()
+def wechat_get_article_detail(url: str) -> Dict[str, Any]:
+    """根据文章链接获取公众号文章详情，支持全文内容（优质库）。
+    url 形如 https://mp.weixin.qq.com/s/..."""
+    return _call(lambda: _get_client().wechat.get_article_detail, url=url)
+
+
+@mcp.tool()
+def wechat_search_ai_articles(keyword: str, page_num: int = 1, page_size: int = 20,
+                              start_time: Optional[str] = None,
+                              end_time: Optional[str] = None) -> Dict[str, Any]:
+    """搜索公众号 AI 创作相关文章（优质库）。keyword 必填；
+    start_time/end_time 格式如 "2026-06-01 00:00:00"。"""
+    return _call(lambda: _get_client().wechat.search_ai_articles,
+                 keyword=keyword, page_num=page_num, page_size=page_size,
+                 start_time=start_time, end_time=end_time)
+
+
+# ─── B 站 ────────────────────────────────────────────────
+
+
+@mcp.tool()
+def bilibili_search_articles(keyword: str, page: int = 1,
+                             page_size: Optional[int] = None,
+                             order: Optional[str] = None) -> Dict[str, Any]:
+    """搜索 B 站视频（优质库）。keyword 必填；page 从 1 开始；page_size 默认 10、最大 50；
+    order：time=发布时间 / play=播放数 / like=点赞数 / comment=评论数 / favorite=收藏数。"""
+    return _call(lambda: _get_client().bilibili.search_articles,
+                 keyword=keyword, page=page, page_size=page_size, order=order)
+
+
+@mcp.tool()
+def bilibili_search_users(keyword: str, page: int = 1,
+                          page_size: Optional[int] = None,
+                          order: Optional[str] = None) -> Dict[str, Any]:
+    """搜索 B 站 UP 主（优质库）。keyword 必填；
+    order：follower=粉丝数 / like=获赞数，默认按相关性。"""
+    return _call(lambda: _get_client().bilibili.search_users,
+                 keyword=keyword, page=page, page_size=page_size, order=order)
+
+
+@mcp.tool()
+def bilibili_get_account(mid: str) -> Dict[str, Any]:
+    """获取 B 站 UP 主信息（优质库）。mid 为 B 站用户唯一 ID。"""
+    return _call(lambda: _get_client().bilibili.get_account, mid=mid)
+
+
+@mcp.tool()
+def bilibili_get_user_works(mid: Optional[str] = None,
+                            account_url: Optional[str] = None, page: int = 1,
+                            page_size: Optional[int] = None,
+                            order: Optional[str] = None) -> Dict[str, Any]:
+    """获取 B 站 UP 主视频列表（优质库）。mid 与 account_url（主页链接）至少传一个；
+    order：time=发布时间 / play=播放数 / like=点赞数。"""
+    return _call(lambda: _get_client().bilibili.get_user_works,
+                 mid=mid, account_url=account_url, page=page,
+                 page_size=page_size, order=order)
+
+
+@mcp.tool()
+def bilibili_get_work(bvid: Optional[str] = None,
+                      work_url: Optional[str] = None) -> Dict[str, Any]:
+    """获取 B 站视频详情（优质库）。bvid 与 work_url 至少传一个，
+    work_url 支持 bilibili.com/video/BVxxx 或 b23.tv 短链。"""
+    return _call(lambda: _get_client().bilibili.get_work, bvid=bvid, work_url=work_url)
+
+
+# ─── 今日头条 ────────────────────────────────────────────
+
+
+@mcp.tool()
+def toutiao_search_works(keyword: str, offset: int = 0) -> Dict[str, Any]:
+    """搜索今日头条内容（实时）。keyword 必填；offset 翻页偏移从 0 开始、每页 +1。"""
+    return _call(lambda: _get_client().toutiao.search_works, keyword=keyword, offset=offset)
+
+
+@mcp.tool()
+def toutiao_get_work(opus_id: str) -> Dict[str, Any]:
+    """获取今日头条作品详情（实时）。opus_id 为作品 ID。"""
+    return _call(lambda: _get_client().toutiao.get_work, opus_id=opus_id)
+
+
+# ─── TikTok ──────────────────────────────────────────────
+
+
+@mcp.tool()
+def tiktok_search_users(keyword: str, cursor: int = 0,
+                        rid: Optional[str] = None) -> Dict[str, Any]:
+    """搜索 TikTok 账号。keyword 必填；cursor 翻页游标第一页为 0、每页 +10；
+    rid 为上一页数据返回的 rid，翻页时传入。返回含 cursor/hasMore/userList。"""
+    return _call(lambda: _get_client().tiktok.search_users,
+                 keyword=keyword, cursor=cursor, rid=rid)
+
+
+# ─── AI 搜索（自动轮询）──────────────────────────────────
+
+
+@mcp.tool()
+def ai_search_kimi(inquiry_text: str, timeout_seconds: int = 240) -> Dict[str, Any]:
+    """Kimi 联网 AI 搜索，提交后自动等待并返回完整结果（含 content/webPages）。
+    inquiry_text 为搜索提问文本；超时未完成时返回 taskId，可用 ai_search_kimi_result 再查。"""
+    return _run_task(lambda: _get_client().ai_search.kimi_submit,
+                     lambda: _get_client().ai_search.kimi_result,
+                     timeout_seconds, inquiry_text=inquiry_text)
+
+
+@mcp.tool()
+def ai_search_doubao(inquiry_text: str, timeout_seconds: int = 240) -> Dict[str, Any]:
+    """豆包联网 AI 搜索，提交后自动等待并返回完整结果。
+    inquiry_text 为搜索提问文本；超时未完成时返回 taskId，可用 ai_search_doubao_result 再查。"""
+    return _run_task(lambda: _get_client().ai_search.doubao_submit,
+                     lambda: _get_client().ai_search.doubao_result,
+                     timeout_seconds, inquiry_text=inquiry_text)
+
+
+@mcp.tool()
+def ai_search_deepseek(inquiry_text: str, timeout_seconds: int = 240) -> Dict[str, Any]:
+    """Deepseek 联网 AI 搜索，提交后自动等待并返回完整结果。
+    inquiry_text 为搜索提问文本；超时未完成时返回 taskId，可用 ai_search_deepseek_result 再查。"""
+    return _run_task(lambda: _get_client().ai_search.deepseek_submit,
+                     lambda: _get_client().ai_search.deepseek_result,
+                     timeout_seconds, inquiry_text=inquiry_text)
+
+
+@mcp.tool()
+def ai_search_kimi_result(task_id: str) -> Dict[str, Any]:
+    """查询 Kimi 搜索任务结果。仅在 ai_search_kimi 超时返回 taskId 后使用。"""
+    return _call(lambda: _get_client().ai_search.kimi_result, task_id=task_id)
+
+
+@mcp.tool()
+def ai_search_doubao_result(task_id: str) -> Dict[str, Any]:
+    """查询豆包搜索任务结果。仅在 ai_search_doubao 超时返回 taskId 后使用。"""
+    return _call(lambda: _get_client().ai_search.doubao_result, task_id=task_id)
+
+
+@mcp.tool()
+def ai_search_deepseek_result(task_id: str) -> Dict[str, Any]:
+    """查询 Deepseek 搜索任务结果。仅在 ai_search_deepseek 超时返回 taskId 后使用。"""
+    return _call(lambda: _get_client().ai_search.deepseek_result, task_id=task_id)
+
+
+# ─── AI 生成（自动轮询）──────────────────────────────────
+
+
+@mcp.tool()
+def gpt_image_generate(prompt: str, n: int = 1, size: str = "1024x1024",
+                       quality: str = "medium", background: str = "auto",
+                       output_format: str = "png",
+                       output_compression: Optional[int] = None,
+                       model_name: str = "gpt-image-2",
+                       operation: str = "generate",
+                       input_fidelity: Optional[str] = None,
+                       images: Optional[List[Dict[str, str]]] = None,
+                       timeout_seconds: int = 240) -> Dict[str, Any]:
+    """GPT 图片生成（gpt-image-2），提交后自动等待并返回 imagePaths。
+    operation：generate=文生图 / edit=图生图（edit 时 images 必填，
+    形如 [{"url": "https://..."}]，input_fidelity 支持 high/low）；
+    quality：low/medium/high/auto；background：transparent/opaque/auto；
+    output_format：png/jpeg/webp。超时返回 taskId，可用 gpt_image_result 再查。"""
+    return _run_task(lambda: _get_client().gpt_image.submit,
+                     lambda: _get_client().gpt_image.result,
+                     timeout_seconds, prompt=prompt, n=n, size=size, quality=quality,
+                     background=background, output_format=output_format,
+                     output_compression=output_compression, model_name=model_name,
+                     operation=operation, input_fidelity=input_fidelity, images=images)
+
+
+@mcp.tool()
+def gpt_image_result(task_id: str) -> Dict[str, Any]:
+    """查询 GPT 图片生成任务结果（含 status/imagePaths/failReason）。
+    仅在 gpt_image_generate 超时返回 taskId 后使用。"""
+    return _call(lambda: _get_client().gpt_image.result, task_id=task_id)
+
+
+@mcp.tool()
+def doubao_image_pro_generate(prompt: str, size: str = "2048x2048",
+                              image: Optional[Any] = None,
+                              output_format: str = "jpeg",
+                              response_format: str = "url",
+                              watermark: bool = True,
+                              optimize_prompt: bool = False,
+                              optimize_mode: str = "standard",
+                              timeout_seconds: int = 240) -> Dict[str, Any]:
+    """豆包 Seedream 5.0 Pro 图片生成，提交后自动等待结果。
+    size 支持 "1K"/"2K" 或像素值如 "2048x2048"；image 为图生图输入（URL 或 Base64，
+    可传单个字符串或字符串列表）；optimize_prompt 开启提示词优化（standard/fast）。
+    超时返回 taskId，可用 doubao_image_pro_result 再查。"""
+    return _run_task(lambda: _get_client().doubao_image.pro_submit,
+                     lambda: _get_client().doubao_image.pro_result,
+                     timeout_seconds, prompt=prompt, size=size, image=image,
+                     output_format=output_format, response_format=response_format,
+                     watermark=watermark, optimize_prompt=optimize_prompt,
+                     optimize_mode=optimize_mode)
+
+
+@mcp.tool()
+def doubao_image_pro_result(task_id: str) -> Dict[str, Any]:
+    """查询 Seedream 5.0 Pro 任务结果。仅在 doubao_image_pro_generate 超时返回 taskId 后使用。"""
+    return _call(lambda: _get_client().doubao_image.pro_result, task_id=task_id)
+
+
+@mcp.tool()
+def doubao_image_lite_generate(prompt: str, size: str = "2048x2048",
+                               image: Optional[Any] = None,
+                               output_format: str = "jpeg",
+                               response_format: str = "url",
+                               watermark: bool = True,
+                               sequential: str = "disabled",
+                               max_images: int = 4,
+                               optimize_prompt: bool = False,
+                               optimize_mode: str = "standard",
+                               timeout_seconds: int = 240) -> Dict[str, Any]:
+    """豆包 Seedream 5.0 Lite 图片生成，支持组图，提交后自动等待结果。
+    size 支持 "2K"/"3K"/"4K" 或像素值；sequential 设为 "auto" 时生成组图，
+    max_images 控制组图数量（1~10）。超时返回 taskId，可用 doubao_image_lite_result 再查。"""
+    return _run_task(lambda: _get_client().doubao_image.lite_submit,
+                     lambda: _get_client().doubao_image.lite_result,
+                     timeout_seconds, prompt=prompt, size=size, image=image,
+                     output_format=output_format, response_format=response_format,
+                     watermark=watermark, sequential=sequential, max_images=max_images,
+                     optimize_prompt=optimize_prompt, optimize_mode=optimize_mode)
+
+
+@mcp.tool()
+def doubao_image_lite_result(task_id: str) -> Dict[str, Any]:
+    """查询 Seedream 5.0 Lite 任务结果。仅在 doubao_image_lite_generate 超时返回 taskId 后使用。"""
+    return _call(lambda: _get_client().doubao_image.lite_result, task_id=task_id)
+
+
+@mcp.tool()
+def doubao_video_generate(content: List[Dict[str, Any]],
+                          model: Optional[str] = None,
+                          resolution: str = "720p",
+                          ratio: str = "adaptive",
+                          duration: int = 5,
+                          seed: int = -1,
+                          watermark: bool = False,
+                          generate_audio: bool = True,
+                          return_last_frame: bool = False,
+                          timeout_seconds: int = 480) -> Dict[str, Any]:
+    """豆包 Seedance 2.0 视频生成，提交后自动等待结果。
+    content 为输入内容列表（必填），每项形如：
+    {"type": "text", "text": "描述文字"}；
+    {"type": "image_url", "imageUrl": "图片URL", "imageRole": "first_frame/reference_image"}；
+    {"type": "video_url", "videoUrl": "视频URL", "videoRole": "reference_video"}；
+    {"type": "audio_url", "audioUrl": "音频URL", "audioRole": "reference_audio"}。
+    resolution：480p/720p/1080p；ratio：adaptive/16:9/4:3/1:1/3:4/9:16/21:9；
+    duration 秒数 [4,15] 或 -1 智能。超时返回 taskId，可用 doubao_video_result 再查。"""
+    return _run_task(lambda: _get_client().doubao_video.submit,
+                     lambda: _get_client().doubao_video.result,
+                     timeout_seconds, content=content, model=model,
+                     resolution=resolution, ratio=ratio, duration=duration,
+                     seed=seed, watermark=watermark, generate_audio=generate_audio,
+                     return_last_frame=return_last_frame)
+
+
+@mcp.tool()
+def doubao_video_result(task_id: str) -> Dict[str, Any]:
+    """查询豆包视频生成任务结果。仅在 doubao_video_generate 超时返回 taskId 后使用。"""
+    return _call(lambda: _get_client().doubao_video.result, task_id=task_id)
+
+
+def main() -> None:
+    """以 stdio transport 启动 MCP server"""
+    mcp.run()
+
+
+if __name__ == "__main__":
+    main()