前言
MCP（Model Context Protocol）在2026年已经不再是"概念"，它正在成为AI工程师构建生产级工具链的核心协议标准。本文聚焦于一个非常具体的工程场景：如何利用MCP协议，将亚马逊实时数据接入AI工作流，实现亚马逊MCP数据运营的完整闭环。
适合人群：AI应用工程师、跨境电商技术负责人、对MCP协议感兴趣的开发者。
MCP协议核心原理
MCP采用JSON-RPC 2.0作为底层传输协议，支持stdio和HTTP+SSE两种传输方式。其最核心的设计决策是Server-driven Tool Declaration：工具的能力描述由Server端声明，Host/Client无需预先知道可用工具的列表，一切在运行时动态协商。
这个设计让"为AI配备实时数据工具"这件事变得极度灵活。我们只需要把Pangolinfo的亚马逊数据采集能力封装成符合MCP规范的Server，AI模型就可以在任何支持MCP的Host环境中调用它。
工具暴露规范
{
"name": "scan_bsr_opportunities",
"description": "实时扫描亚马逊指定类目BSR机会品，基于Pangolinfo分钟级数据更新",
"inputSchema": {
"type": "object",
"properties": {
"category_id": {"type": "string", "description": "亚马逊类目节点ID"},
"marketplace": {"type": "string", "default": "US"},
"max_reviews": {"type": "integer", "default": 150},
"days": {"type": "integer", "default": 7}
},
"required": ["category_id"]
}
}
系统架构设计
┌─────────────────────────────────────────────────┐
│ MCP Host │
│ ┌──────────────┐ ┌──────────────────────┐ │
│ │ AI Model │◄──►│ MCP Client │ │
│ │ (Claude) │ │ (Session Manager) │ │
│ └──────────────┘ └──────────┬───────────┘ │
└────────────────────────────────┼────────────────┘
│ JSON-RPC (stdio/SSE)
┌────────────────────────────────▼────────────────┐
│ MCP Server │
│ ┌────────────────────────────────────────────┐ │
│ │ Tool Registry │ │
│ │ scan_bsr_opportunities() │ │
│ │ get_product_reviews_analysis() │ │
│ │ get_ad_placement_intelligence() │ │
│ └────────────────┬───────────────────────────┘ │
└───────────────────┼─────────────────────────────┘
│ HTTP REST
┌───────────────────▼─────────────────────────────┐
│ Pangolinfo Data Layer │
│ ┌────────────────┐ ┌──────────────────────┐ │
│ │ Scrape API │ │ Reviews Scraper API │ │
│ │ (BSR/榜单/ │ │ (评论/Customer Says) │ │
│ │ 广告位数据) │ └──────────────────────┘ │
│ └────────────────┘ │
│ 分钟级更新 | SP广告位98%采集率 | 结构化JSON输出 │
└─────────────────────────────────────────────────┘
│ 亚马逊平台公开数据采集
┌────────▼────────────────────────────────────────┐
│ Amazon Marketplace │
│ BSR榜单 | New Releases | ABA搜索词 | SP广告位 │
└─────────────────────────────────────────────────┘
核心实现
MCP Server（工具暴露层）
# amazon_mcp_server.py
import asyncio
import json
import requests
import os
from mcp.server import Server
from mcp.server.stdio import stdio_server
from mcp.server.models import InitializationOptions
from mcp import types
server = Server("pangolinfo-amazon")
API_KEY = os.getenv("PANGOLINFO_API_KEY")
BASE = "https://api.pangolinfo.com/v2"
@server.list_tools()
async def list_tools():
return [
types.Tool(
name="scan_bsr_opportunities",
description="实时扫描亚马逊类目BSR机会品（Pangolinfo分钟级数据）",
inputSchema={
"type": "object",
"properties": {
"category_id": {"type": "string"},
"marketplace": {"type": "string", "default": "US"},
"days": {"type": "integer", "default": 7},
"max_reviews": {"type": "integer", "default": 150},
"min_sales": {"type": "integer", "default": 300}
},
"required": ["category_id"]
}
),
types.Tool(
name="analyze_product_reviews",
description="获取ASIN差评数据并分析产品迭代优先级（含Customer Says摘要）",
inputSchema={
"type": "object",
"properties": {
"asin": {"type": "string"},
"marketplace": {"type": "string", "default": "US"},
"focus": {"type": "string", "default": "negative"}
},
"required": ["asin"]
}
)
]
@server.call_tool()
async def call_tool(name: str, arguments: dict):
headers = {"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"}
if name == "scan_bsr_opportunities":
resp = requests.post(f"{BASE}/amazon/bestsellers", json={
"category_id": arguments["category_id"],
"country": arguments.get("marketplace", "US"),
"days": arguments.get("days", 7),
"filters": {
"max_reviews": arguments.get("max_reviews", 150),
"min_monthly_sales": arguments.get("min_sales", 300)
}
}, headers=headers, timeout=30)
data = resp.json()
result = {
"opportunities_count": len(data.get("results", [])),
"top_opportunities": [{
"asin": p["asin"],
"title": p["title"][:60] + "...",
"bsr_rank": p["bsr_current"],
"bsr_change": f"+{p['bsr_change']}",
"reviews": p["review_count"],
"monthly_sales": p["monthly_sales"],
"price": p["price"]
} for p in data.get("results", [])[:10]]
}
elif name == "analyze_product_reviews":
resp = requests.post(f"{BASE}/amazon/reviews", json={
"asin": arguments["asin"],
"country": arguments.get("marketplace", "US"),
"rating_filter": "critical",
"include_customer_says": True,
"max_pages": 10
}, headers=headers, timeout=60)
data = resp.json()
result = {
"asin": arguments["asin"],
"customer_says_summary": data.get("customer_says", ""),
"negative_themes": data.get("negative_themes", [])[:5],
"improvement_priorities": data.get("improvement_suggestions", [])
}
else:
result = {"error": f"Unknown tool: {name}"}
return [types.TextContent(type="text", text=json.dumps(result, ensure_ascii=False, indent=2))]
async def main():
async with stdio_server() as (r, w):
await server.run(r, w, InitializationOptions(
server_name="pangolinfo-amazon",
server_version="1.0.0"
))
if __name__ == "__main__":
asyncio.run(main())
工作流客户端
# workflow_client.py
import asyncio
import json
import anthropic
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client
async def run_selection_analysis(category_id: str, marketplace: str = "US"):
"""完整的亚马逊MCP数据运营工作流"""
params = StdioServerParameters(
command="python",
args=["amazon_mcp_server.py"]
)
client = anthropic.Anthropic()
async with stdio_client(params) as (r, w):
async with ClientSession(r, w) as session:
await session.initialize()
# 获取工具列表
tools_resp = await session.list_tools()
tools = [{
"name": t.name,
"description": t.description,
"input_schema": t.inputSchema
} for t in tools_resp.tools]
messages = [{
"role": "user",
"content": f"""
请对亚马逊 {marketplace} 站 {category_id} 类目做一次完整的机会分析：
1. 扫描过去7天的BSR机会品（评论<150，月销>300）
2. 对评分最高的机会品分析其差评和改进方向
3. 给出3个最值得关注的机会品推荐 + 切入建议
"""
}]
# AI 推理循环
while True:
response = client.messages.create(
model="claude-opus-4-5",
max_tokens=4096,
tools=tools,
messages=messages
)
if response.stop_reason == "end_turn":
return next((b.text for b in response.content if hasattr(b, "text")), "")
if response.stop_reason == "tool_use":
tool_results = []
for block in response.content:
if block.type == "tool_use":
result = await session.call_tool(block.name, block.input)
tool_results.append({
"type": "tool_result",
"tool_use_id": block.id,
"content": result.content[0].text
})
messages.extend([
{"role": "assistant", "content": response.content},
{"role": "user", "content": tool_results}
])
if __name__ == "__main__":
result = asyncio.run(run_selection_analysis("2975312011"))
print(result)
最佳实践
数据缓存策略
import redis
import hashlib
cache = redis.Redis()
def cached_api_call(endpoint: str, payload: dict, ttl: int = 300):
"""5分钟TTL缓存，避免重复请求相同类目数据"""
cache_key = hashlib.md5(f"{endpoint}{json.dumps(payload, sort_keys=True)}".encode()).hexdigest()
cached = cache.get(cache_key)
if cached:
return json.loads(cached)
result = requests.post(endpoint, json=payload, headers=get_headers()).json()
cache.setex(cache_key, ttl, json.dumps(result))
return result
错误处理最佳实践
from tenacity import retry, stop_after_attempt, wait_exponential, retry_if_exception_type
@retry(
stop=stop_after_attempt(3),
wait=wait_exponential(multiplier=1, min=2, max=10),
retry=retry_if_exception_type(requests.HTTPError)
)
def resilient_api_call(url, payload, headers):
resp = requests.post(url, json=payload, headers=headers, timeout=30)
resp.raise_for_status()
return resp.json()
总结
本文展示了基于MCP协议的亚马逊数据运营工作流的完整工程实现。核心设计要点：
- MCP Server负责将Pangolinfo API能力封装为标准化工具，保证AI可动态发现和调用
- 工具描述的质量直接影响AI选择工具的准确性，description字段要写得具体
- 数据质量是整个系统的上限：Pangolinfo的分钟级更新和结构化JSON输出是让AI推理有意义的前提
- 生产环境建议加入缓存层和重试机制，保证工作流稳定性
完整代码已整理完毕。如需AMZ Data Tracker可视化版本的集成方案，欢迎在评论区留言交流。
