Chinese e-commerce MCP servers — let AI agents read merchant business data (Ocean Engine, Douyin Shop, JD.com, Taobao, Pinduoduo). 电商经营数据 MCP 连接器。
mcp-cn-commerce — Chinese E-Commerce Platform MCP Servers
🛒 Let AI agents read Chinese e-commerce merchant business data. Not content publishing — business operations MCP connectors for AI agents like Claude, ChatGPT, Gemini.
Keywords: MCP Server, Model Context Protocol, Chinese e-commerce, AI agent, e-commerce data, 电商 MCP, 抖店, 京东, 巨量引擎, Ocean Engine, Douyin Shop, JD.com, Taobao, Pinduoduo, Python MCP, AI business intelligence, agent tool
English | 简体中文
Table of Contents
- What is mcp-cn-commerce?
- Why This Project
- Supported Platforms
- Quick Start
- Architecture
- Tools Reference
- Security
- Docker - Docker 部署与配置
- Examples - 使用示例和场景
- FAQ
- Contributing
- Roadmap
What is mcp-cn-commerce?
A monorepo of independent MCP (Model Context Protocol) servers that give AI agents structured, type-safe access to Chinese e-commerce platform business data. Each server wraps one platform's open API:
- 巨量引擎 (Ocean Engine) — advertising campaign, report, and account data
- 抖店 (Douyin Shop) — orders, products, refunds, shop management
- 京东 (JD.com) — orders, products, shop information
- More coming: 淘宝, 拼多多, 快手, 小红书, 微信小店
All tools are read-only by default — AI agents can analyze your business data but cannot modify anything.
Why This Project
Existing Chinese-platform MCP servers all focus on content publishing (posting videos, searching trends). Zero cover merchant business operations — pulling ad reports, checking orders, managing refunds.
| | Content MCPs (HuiMei, Astron, etc.) | mcp-cn-commerce | |---|---|---| | Purpose | Post videos, search hot topics | Pull ad reports, query orders | | Target User | Creators / influencers | E-commerce business owners / operators | | Data Type | Content data (views, likes, trends) | Business data (revenue, orders, refunds, ROAS) | | Platform Scope | Content platforms | E-commerce + advertising platforms | | Operations | Publish / write | Read-only analytics & monitoring |
This is the first open-source MCP server suite for Chinese e-commerce business operations.
Platforms
| Platform | Category | Phase | Status | Tests | Open API | |---|---|---|---|---|---| | 巨量引擎 (Ocean Engine) | Advertising (广告投放) | 1 | ✅ | 24 | open.oceanengine.com | | 巨量千川 (Qianchuan) | E-commerce Ads (电商广告) | 1 | ✅ | (shared) | qianchuan.jinritemai.com | | 抖店 (Douyin Shop) | E-commerce (电商店铺) | 1 | ✅ | 31 | op.jinritemai.com | | 京东 (JD.com) | E-commerce (电商店铺) | 1 | ✅ | 19 | jos.jd.com | | 淘宝 (Taobao) | E-commerce (电商店铺) | 2 | ✅ | 36 | open.taobao.com | | 拼多多 (Pinduoduo) | E-commerce (电商店铺) | 2 | ✅ | 30 | open.pinduoduo.com | | 快手 (Kuaishou) | E-commerce (电商店铺) | 3 | ✅ | 33 | open.kuaixiaodian.com | | 小红书 (Xiaohongshu) | E-commerce (电商店铺) | 3 | ✅ | 33 | open.xiaohongshu.com | | 微信小店 (WeChat Store) | E-commerce (电商店铺) | 3 | ✅ | 25 | developers.weixin.qq.com |
Phase 4 (exploratory): 闲鱼 (Xianyu), 美团 (Meituan), 饿了么 (Ele.me) — restricted APIs, awaiting policy clarity.
358 tests across all 8 servers. CI runs on Python 3.11, 3.12, 3.13.
Quick Start
Docker (recommended — no local Python setup)
# Build the image
docker build -t mcp-cn-commerce .
# Run tests
docker run --rm mcp-cn-commerce make test
# Run a platform server (Ocean Engine example)
docker run --rm -i --env-file .env mcp-cn-commerce mcp-cn-oceanengine
See Docker documentation for full usage, MCP client configuration, and Docker Compose shortcuts.
Installation
From GitHub Releases (recommended — no registration needed)
# Visit the latest Release and download the .whl file
# https://github.com/TonyWang-hub/mcp-cn-commerce/releases/latest
# Or install directly from the Release URL:
pip install https://github.com/TonyWang-hub/mcp-cn-commerce/releases/latest/download/mcp_cn_commerce-0.1.0-py3-none-any.whl
From Git (always latest)
pip install git+https://github.com/TonyWang-hub/mcp-cn-commerce.git
For development
git clone https://github.com/TonyWang-hub/mcp-cn-commerce.git
cd mcp-cn-commerce
pip install -e ".[dev]"
Configuration
Set credentials via environment variables:
# 巨量引擎 (Ocean Engine) — Advertising platform
export OCEANENGINE_APP_KEY="your_app_key"
export OCEANENGINE_APP_SECRET="your_app_secret"
export OCEANENGINE_ACCESS_TOKEN="your_access_token"
# 抖店 (Douyin Shop) — TikTok Shop China
export DOUDIAN_APP_KEY="your_app_key"
export DOUDIAN_APP_SECRET="your_app_secret"
export DOUDIAN_SHOP_ID="your_shop_id"
export DOUDIAN_ACCESS_TOKEN="your_access_token"
# 京东 (JD.com) — Jingdong e-commerce
export JD_APP_KEY="your_app_key"
export JD_APP_SECRET="your_app_secret"
export JD_ACCESS_TOKEN="your_access_token"
Add to Your MCP Client
Works with Claude Desktop, Cherry Studio, Kimi Work, and any MCP-compatible AI client:
{
"mcpServers": {
"oceanengine": {
"command": "mcp-cn-oceanengine"
},
"doudian": {
"command": "mcp-cn-doudian"
},
"jd": {
"command": "mcp-cn-jd"
}
}
}
Example: AI Agent Querying Your Business
Once connected, you can ask your AI agent questions like:
"Show me this week's Ocean Engine campaign ROAS, sorted by spend" "Which Douyin Shop products are low on stock?" "How many JD refunds are pending approval?" "Compare my ad performance across Ocean Engine campaigns this month vs last month"
Architecture
mcp-cn-commerce/
├── .github/workflows/test.yml # CI: pytest on push (3.11/3.12/3.13)
├── shared/ # Shared auth/signing/pagination
│ └── cn_commerce_base.py # CommerceMCPBase — extend for new platforms
├── servers/
│ ├── oceanengine/ 17 tools ├── doudian/ 20 tools
│ ├── jd/ 14 tools ├── taobao/ 13 tools
│ ├── pinduoduo/ 13 tools ├── kuaishou/ 12 tools
│ ├── xiaohongshu/ 12 tools └── weixin-store/ 11 tools
├── docs/platforms.md # 8-platform API comparison & auth matrix
├── docs/docker.md # Docker deployment guide
├── Dockerfile # Multi-platform MCP server image
├── docker-compose.yml # Local development shortcuts
├── README.md / README_zh.md # English / 简体中文
└── LICENSE # MIT
Each servers/<platform>/ is an independent MCP server. Users install only what they need.
Tools Summary
| Server | Tools | Categories | |---|---|---| | oceanengine | 17 | Ads, Qianchuan, Star, Creative, Audience, Optimization | | doudian | 20 | Orders, Products, Refunds, Logistics, Reviews, Live, Traffic, Marketing, Billing, Shop | | jd | 14 | Orders, Products, After-Sale, Logistics, Reviews, Pricing, Inventory, Marketing, Shop | | taobao | 13 | Orders, Products, Refunds, Logistics, Reviews, Shop, Marketing, Categories | | pinduoduo | 13 | Orders, Products, Refunds, Logistics, Reviews, Shop, Marketing, Affiliate | | kuaishou | 12 | Orders, Products, Refunds, Logistics, Reviews, Shop, Marketing | | xiaohongshu | 12 | Orders, Products, Refunds, Logistics, Reviews, Shop, Marketing, Inventory, Finance | | weixin-store | 11 | Orders, Products, Refunds, Logistics, Shop, Marketing, Supply Chain, Categories | | Total | 112 | All read-only by default |
For full tool details, see the source code in each servers/<platform>/src/ directory.
| get_product_list | Product catalog with pricing and stock | /product/list |
| get_shop_info | Merchant shop information | /shop/info |
Security
This project handles sensitive e-commerce API credentials. Our security guarantees:
- 🔒 Runs locally — API keys and secrets never leave your machine
- 📖 Open source — every line of code is auditable
- 👁️ Read-only by default — all 112 tools only read data; zero write/modify/delete operations
- 📡 No telemetry — no usage data is collected, tracked, or transmitted
- 🖥️ Direct API calls — connects directly to platform APIs; no intermediate server or proxy
- 🔑 Env-var config — credentials are loaded from environment variables, never hardcoded
Roadmap
Phase 1 — Foundation ✅
- 巨量引擎: Ad campaign & report read APIs
- 巨量千川: E-commerce advertising (shared Ocean Engine auth)
- 抖店: Order, product, after-sale read APIs
- 京东: Order, product, shop read APIs
Phase 2 — Mid-Tier Expansion ✅
- 淘宝 (Taobao): Full Top API integration — orders, products, logistics
- 拼多多 (Pinduoduo): Orders, products, promotion tools
Phase 3 — Long-Tail Coverage ✅
- 快手 (Kuaishou): Orders, products, logistics
- 小红书 (Xiaohongshu): Orders, products, inventory
- 微信小店 (WeChat Store): Orders, products, after-sale
Phase 4 — Exploratory ⬜
- 闲鱼, 美团, 饿了么 (API access pending policy)
Related Resources
- Model Context Protocol (MCP) Documentation
- MCP Python SDK
- Claude Desktop — MCP Support
- Cherry Studio — Multi-Model MCP Client
- Platform API Comparison
Citation
If you use mcp-cn-commerce in your research or project:
@software{mcp-cn-commerce,
author = {Wang, Zhuo},
title = {mcp-cn-commerce: MCP Servers for Chinese E-Commerce Platforms},
year = {2026},
url = {https://github.com/TonyWang-hub/mcp-cn-commerce}
}
Star History
License
MIT — see LICENSE.