Skip to main content
Glama
Eureka-Zhang

airus-mcp

by Eureka-Zhang
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Local

airus-mcp

1688 / OZON 电商爬虫的 MCP Server。把选品爬虫暴露为标准 MCP 工具，供 Claude 桌面版、Cowork、Claude Code 等任何 MCP 宿主直接调用——agent 当大脑，爬虫当手。

由 AIRus 项目的爬虫核心迁移而来，剥离了飞书机器人层与自建 WebSearch 层（宿主 agent 的原生联网能力已覆盖），并修复了迁移前的若干已知 bug。

架构

MCP 宿主（Claude Desktop / Cowork / Claude Code …）
    │  stdio
    ▼
mcp_server.py            ← MCP 工具入口（本仓库）
    │  HTTP :8765
    ▼
web_app.py               ← 任务队列 + 爬虫调度（单账号串行）
    │  subprocess
    ├── scrape_1688.py   ← 1688 关键词搜索 / 以图搜款 / 店铺联系方式（需登录态）
    └── crawl_ozon.py    ← OZON 关键词搜索（需俄罗斯 IP）
         ▲
         └── Chrome CDP :9222（真实 Chrome，登录态 / 过反爬的根基）

Related MCP server: wildberries-mcp

MCP 工具（7 个）

工具

说明

crawler_browser_status

检查 Chrome CDP 与队列状态，是否需要人工登录/验证码

crawler_create_job

创建异步采集任务（1688 / ozon）

crawler_image_search

1688 以图搜款：传本地图片路径，自动上传并建任务

crawler_wait_job

阻塞等待任务完成（推荐，代替客户端轮询）

crawler_get_job

查询任务状态与结果（含中文字段的 bitable_rows）

crawler_list_jobs

列出近期任务

crawler_cancel_job

取消排队中的任务

快速开始

1. 启动爬虫 Worker（Mac 采集机）

./start_mac_worker.sh

脚本会自动创建 venv、装依赖、拉起 Chrome CDP (:9222) 和 web_app (:8765)。 首次使用请在弹出的 Chrome 窗口里手动登录 1688（登录态保存在 ~/.1688-tool-chrome）。

2. 注册 MCP Server

Claude Desktop 的 claude_desktop_config.json（或任何 MCP 宿主的等价配置）：

{
  "mcpServers": {
    "airus-crawler": {
      "command": "/path/to/airus-mcp/.venv/bin/python",
      "args": ["/path/to/airus-mcp/mcp_server.py"],
      "env": {
        "CRAWLER_BASE_URL": "http://127.0.0.1:8765"
      }
    }
  }
}

3. 对话示例

帮我在 1688 搜「蓝牙耳机」前 2 页，采集供应商联系方式
帮我爬 OZON 上「механическая клавиатура」的第 1 页
用这张图在 1688 找同款：/Users/me/Desktop/product.jpg

环境变量

变量

默认

说明

CRAWLER_BASE_URL

http://127.0.0.1:8765

web_app 地址（MCP 端）

WEB_HOST / WEB_PORT

127.0.0.1 / 8765

web_app 监听地址（Worker 端）

RUB_CNY_RATE

0.078

卢布→人民币汇率（OZON 价格换算）

CDP_PORT

9222

Chrome CDP 端口

COMPANY_WORKER_URL

—

设置后 web_app 变为到远端 Worker 的反向代理

CRAWLER_WAIT_POLL_INTERVAL

10

crawler_wait_job 轮询间隔（秒）

注意事项

  • 1688：必须已登录（诚信通账号）；出现滑块/验证码时 browser_status.needs_action 会提示，请到采集机 Chrome 窗口人工完成。

  • OZON：采集机必须是俄罗斯出口 IP；无需登录。

  • 任务队列为单账号串行（物理上只有一个 Chrome），并发请求会自动排队。

  • 队列当前为内存态，web_app 重启后任务与结果丢失（结果同时落盘在 scrape_result/ 的 CSV/JSONL 里）。

相对原项目（AIRus）的变更

  • 删除：飞书机器人层（Node/TS）、webfetch 自建搜索抓取、market_research 双市场调研、同步阻塞的 crawler_search 工具

  • 新增：crawler_wait_job（阻塞等待）、crawler_image_search（本地图片以图搜款）

  • 修复：联系方式按下标兜底导致串店、供应商电话列被覆盖丢失、rating/review_count 字段映射重复定义、汇率硬编码（改为 RUB_CNY_RATE）、多处乱码提示文案、登录误报 marker、删除 3 个遗留死函数

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityStale
ResponsivenessNo issues

Resources

Related MCP Connectors

  • Agentic ShelfA
    ai.agenticshelf

    Hosted MCP for e-commerce: live product catalog, stock, and pricing for AI agents.

    E-commerce & RetailSearch
    2
  • CraloOAuth
    ai.cralo

    Pay-per-use tool marketplace for AI agents. Search, price-check, and call APIs via MCP.

    Agent OrchestrationDeveloper Tools
    1
  • AIsa Shopping & MarketplaceAOAuth
    one.aisa

    Your agent needs marketplace data — what a product costs on Amazon and Google Shopping, who the sellers are, what reviewers actually complain about. **What you can ask for** • "What is this ASIN's price history, rating and seller list?" • "Who else sells this product, and at what price?" • "Pull the reviews for this product and group the complaints." • "What comes up on Google Shopping for this query in the UK?" • "Compare these products across both marketplaces." **How to use it** Point any MCP client at https://mcp.aisa.one/seo-merchant/mcp and sign in with OAuth — there is no key to create or paste. 22 tools: Amazon products, ASIN detail and sellers; Google Shopping products, product info, sellers and reviews; live and queued forms, with raw HTML where you need it. **It is also a door to the rest** The same login reaches 26 sources and 580+ operations. Price the product here, then ask the same agent what the brand's site traffic or ad spend looks like — without adding a second server. **What it costs** Finding and inspecting an operation is free. Running one is billed per call at API prices, with no seat and no monthly minimum, and every call takes max_price_usd so an agent cannot overspend by accident. **Where else it reaches** https://mcp.aisa.one/seo/mcp for all of it at once — rankings, keywords, backlinks, site health and AI-answer visibility across DataForSEO, Semrush and Ahrefs.

    E-commerce & RetailSearch
    271
  • Vee3A
    io.github.Vee3io

    Hosted MCP with 91 agent tools: X, domains, SEO, Maps, Trends, Search, YouTube, TikTok, and more.

    Social MediaBrowser AutomationMarketing
    26921

Related MCP Servers

  • Scavio by scavio-ai

    Scavioofficial

    scavio-ai
    A
    license
    A
    quality
    B
    maintenance
    An MCP server connecting AI agents to Google, YouTube, Amazon, Walmart, TikTok, and Reddit. 21 tools for web search, product lookup, video discovery, and social media analysis.
    21
    322 npm
    7
    MIT

  • wildberries-mcp

    shndo1337
    F
    license
    Not graded
    quality
    C
    maintenance
    MCP server that turns Wildberries marketplace into a toolkit for LLM agents, enabling product search, detailed card inspection, price history, reviews, and cross-product comparison.
    -

  • Scrapingdog MCP Server

    Darshan972
    A
    license
    B
    quality
    B
    maintenance
    MCP server exposing 35 web scraping and SERP tools for general scraping, Google services, e-commerce sites, social media, and search engines. Enables MCP clients like Claude to scrape web pages and search results via natural language.
    35
    28 npm
    MIT