ecommerce-analytical-mcp-server

MCP数据服务

提供方 cyborg1367

数据服务 stdio · 本地进程 未标注 社区 / 第三方 Claude DesktopClaudeCursorVS Code
查看接入方式

A PostgreSQL-backed Model Context Protocol (MCP) server that provides e-commerce data analytics, operations, schema, and safe SQL tools for AI agents.

一句话选型结论

按资料字段整理

ecommerce-analytical-mcp-server 属于数据服务类 MCP 服务器,stdio · 本地进程,鉴权方式未标注,可直接接入 Claude Desktop、Claude、Cursor 等,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类数据服务 传输方式stdio · 本地进程 鉴权未标注 部署方式本地部署
GitHub 星标 1 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

数据服务
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

Local Deployment E-commerce Analytics PostgreSQL Python AI Agent Data Analysis Report Generation

支持客户端

适用场景

Local DeploymentE-commerce AnalyticsPostgreSQL

安装与配置

stdio · 本地进程 未标注
客户端在本机拉起子进程,数据不出本机
uv --directory E:\projects\ecom-mcp run ecom-mcp

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 python,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
数据库查询

用自然语言查库、取数,不用手写 SQL。

数据同步与清洗

把多来源数据拉到一处做整理。

报表取数

按口径快速拉出统计结果。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Claude Desktop、Claude、Cursor,换客户端不用重新研究配置。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在数据服务场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Claude 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
ecommerce-analytical-mcp-server当前 未标注 stdio · 本地进程 本地部署 有 2025-12-17
Memory 9d25f64e 9 未标注 未标注 无 2026-05-17
Getzep MCP 9 stdio · 本地进程 本地部署 有 2026-05-17
anything-llm 未标注 stdio · 本地进程 本地部署 有 2026-04-11
Prisma MCP Server 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Astr Bot 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

数据服务
字段来自项目仓库与公开索引,中文说明由本站标注
提供方cyborg1367上游标注的维护方 cyborg1367
分类数据服务本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式未标注远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · python本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验2025-12-17上游最后核验时间

详细介绍

12.2 KB
来自项目仓库与上游目录的原文,未做删改

E-commerce Analytical MCP Server

A PostgreSQL-backed Model Context Protocol (MCP) server that exposes a curated set of analytics, operations, schema, and safe SQL tools for working with an e-commerce database — so an AI agent (IDE assistant, chat app, etc.) can explore data, generate reports, and produce dashboards without ad-hoc SQL.

Built with FastMCP, SQLAlchemy, and psycopg, with schema reflection to adapt to common variations in real-world e-commerce schemas.


Why this exists

If you’ve ever wanted an AI assistant to reliably answer questions like:

  • “What changed this week?”
  • “Which products are driving revenue?”
  • “Are we building a backlog?”
  • “What should we reorder now?”

…using your actual database (not guesses), this project provides a clean, controlled MCP interface over Postgres.


Key features

  • Analytics tools: Revenue and orders by day, top products, top customers, repeat purchase rate, gross margin, and sales KPIs.
  • Ops and reporting tools: One-page sales and ops reports in Markdown, table counts, low-stock insights, and operational health checks.
  • Visual dashboards: A 2×2 composite sales dashboard PNG (revenue trend, orders trend, top products, KPI tiles).
  • Schema tools: Schema overview, table listing, table descriptions, and schema cache refresh.
  • Safe SQL access: A hardened, read-only SQL tool that only allows SELECT/WITH/SHOW/EXPLAIN.
  • Demo data seeding: Realistic synthetic e-commerce data generator for local experiments.
  • Prompt helpers: Opinionated prompt templates for exec briefs, sales deep dives, ops triage, inventory planning, and data-quality smoke tests.

Architecture overview

  • MCP server: Implemented using fastmcp and started via the ecom-mcp console script.
  • Configuration: Centralized in Settings (Pydantic BaseSettings), loading from environment variables and an optional .env file.
  • Database access: SQLAlchemy with a pooled Engine and a small Unit-of-Work abstraction.
  • Schema reflection: Automatically inspects tables and columns in the public schema to work with common e-commerce schemas (even if column names differ slightly).
  • Services:
  • AnalyticsService: High-level reporting queries (revenue, customers, products, margins, etc.).
  • OpsService: Markdown reports for sales and operations.
  • SchemaService: Introspection helpers for tables and columns.
  • SqlService: Read-only SQL execution with safety checks and timeouts.
  • SeedService: Demo-data writer with safety guardrails.

Requirements

  • Python: >= 3.13
  • PostgreSQL: Any reasonably recent PostgreSQL version
  • Network access: The MCP host must be able to connect to the database provided in the DSN

Quickstart (recommended with uv)```bash

git clone https://github.com/<your-org>/<your-repo>.git
cd <your-repo>

Install dependencies

uv sync

Run the MCP server (stdio)

uv run ecom-mcp


Alternatively:```bash
uv run python -m app.main

Installation (pip)```bash

git clone https://github.com/<your-org>/<your-repo>.git
cd <your-repo>

python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate

pip install .


After installation, the `ecom-mcp` console script will be available on your `PATH`.

---

## Configuration

Configuration is handled by `app.config.settings.Settings`, which reads from:

- environment variables
- an optional `.env` file in the project root

### Main settings

- **`POSTGRES_DSN`** (required): PostgreSQL connection string
  - Accepts `postgresql+psycopg://...`, `postgresql://...`, or `postgres://...`.

- **`ALLOW_WRITES`** (optional, default: `true`): Enables write-capable tools (notably demo data seeding).
  - In production, you typically want this **disabled**.

- **`LOG_LEVEL`** (optional, default: `INFO`): Python logging level (`DEBUG`, `INFO`, `WARNING`, ...)

Example `.env`:```bash
POSTGRES_DSN=postgresql://user:password@localhost:5432/ecom_db
ALLOW_WRITES=false
LOG_LEVEL=INFO

> Note: ALLOW_WRITES=false (or 0) will block tools that modify data, such as seed_demo_data.


Running the MCP server

Start the server via console script:```bash
ecom-mcp


This starts a FastMCP server named **`ecom-postgres`** that communicates over **stdio**, which is the expected transport for most MCP hosts.

Or run the module directly:```bash
python -m app.main

Connecting to MCP hosts

This project runs over stdio by default.

Claude Desktop / Cursor / Trae (JSON with mcpServers)```json

{
"mcpServers": {
"ecom_postgres": {
"command": "uv",
"args": ["--directory", "E:\projects\ecom-mcp", "run", "ecom-mcp"]
}
}
}


### VS Code (create `.vscode/mcp.json`)

VS Code uses a different top-level key (`servers`) and requires `type: &quot;stdio&quot;`:```json
{
  &quot;servers&quot;: {
    &quot;ecomPostgres&quot;: {
      &quot;type&quot;: &quot;stdio&quot;,
      &quot;command&quot;: &quot;uv&quot;,
      &quot;args&quot;: [&quot;--directory&quot;, &quot;${workspaceFolder}&quot;, &quot;run&quot;, &quot;ecom-mcp&quot;],
      &quot;envFile&quot;: &quot;${workspaceFolder}\\.env&quot;
    }
  }
}

MCP tools

Tool identifiers typically match the Python function name.

Health

  • db_ping (health):
    Connectivity check returning current database, user, schema, and server time.

Schema and metadata

  • refresh_schema_cache (schema):
    Clears cached schema metadata. Use after migrations or schema changes.

  • schema_overview (schema):
    Returns a Markdown overview of tables and columns in public, plus structured JSON.

  • list_tables (schema):
    Lists all tables in the public schema.

  • describe_table (schema):
    Returns columns, types, nullability, and default values for a given table.

  • table_counts (schema, debug):
    Row counts for common e-commerce tables (only those that exist).

Analytics

  • revenue_by_day(days=30) (analytics, sales):
    Orders and revenue by day for the last N days (excludes cancelled orders when a status column exists).

  • top_products_last_days(days=30, limit=10) (analytics, sales, catalog):
    Top products by revenue for the last N days.

  • top_customers_last_days(days=90, limit=10) (analytics, sales, customer):
    Top customers by revenue for the last N or 90 days.

  • repeat_purchase_rate(days=180) (analytics, customer):
    Share of customers with 2+ orders over the chosen window.

  • gross_margin_last_days(days=30) (analytics, finance):
    Revenue, cost, gross margin, and margin rate for the last N days.

Operations and reporting

  • low_stock(threshold=10, limit=50) (ops, inventory):
    Lists low-stock items (requires either a v_inventory_on_hand view or an inventory table).

  • ops_health_report(days=14, low_stock_threshold=10) (ops, report):
    Markdown report summarizing order status mix, backlog, and inventory risks.

  • sales_report(days=30, top_n=10) (analytics, report, sales):
    One-page Markdown sales report with KPIs, daily trend, and top products.

Dashboards

  • sales_dashboard(days=30, top_n=10) (analytics, report, charts):
    Generates a 2×2 PNG dashboard (revenue trend, orders trend, top products, KPI tiles).
    Returns both a Markdown description and the PNG image bytes.

> Note: Rendering of tool-returned images depends on the MCP host. Some clients show inline images; others display base64.

SQL (read-only)

  • sql_readonly(query, max_rows=200, timeout_ms=5000) (sql):
    Executes a single read-only SQL statement (SELECT/WITH/SHOW/EXPLAIN only).
    Enforces:
  • Normalized SQL with trailing semicolons stripped
  • No internal semicolons (single statement only)
  • Server-side statement timeout (timeout_ms)
  • Row limit (max_rows)

Seeding and demo data

  • seed_demo_data(size=&quot;small · medium · large&quot;, reset_first=True, seed=42) (seed, demo):
    Inserts realistic demo e-commerce data, optionally truncating existing data first.

  • Requires ALLOW_WRITES=true (or 1) — otherwise raises a permission error.

  • Sizes:
    • small: ~50 customers, 150 products, 600 orders
    • medium: ~300 customers, 900 products, 7,000 orders
    • large: ~1,500 customers, 4,000 products, 40,000 orders

Prompt helpers

The server registers several MCP prompts that guide an AI agent to use tools effectively:

  • weekly_exec_brief(days=7): One-page executive brief (KPIs, what changed, risks, next actions).
  • sales_deep_dive(days=30, top_n=15): Detailed sales analysis (trend, drivers, profitability, recommendations).
  • investigate_revenue_drop(days=14, compare_days=14): Root-cause playbook comparing two windows.
  • ops_triage(days=14): Ops triage focusing on backlog, status mix, and inventory risks.
  • inventory_reorder_plan(days=30, low_stock_threshold=10): Inventory plan using low-stock + sales velocity.
  • data_quality_smoke_test(): Fast sanity checks after seeding or schema changes.

Prompts strongly prefer built-in analytics tools and use sql_readonly only when needed.


Expected schema (high level)

The server is designed to adapt to common e-commerce schemas via reflection, but works best when the following tables (or equivalents) exist in the public schema:

  • customers
  • Recommended columns: customer_id, email, full_name

  • categories (optional but used by the seeder)

  • Recommended columns: category_id, name, slug

  • products

  • Recommended columns: product_id, sku, name, category_id, price
  • Recommended extras: cost, currency_code, attributes, is_active

  • orders

  • Timestamp: one of placed_at, ordered_at, created_at, order_date
  • Total: one of total_amount, grand_total, total
  • Optional: status, order_number, currency_code,
    subtotal_amount, discount_amount, tax_amount, shipping_amount

  • order_items

  • Quantity: quantity or qty
  • Required for seeding: unit_price, sku_snapshot, name_snapshot
  • Optional: unit_cost, line_subtotal, line_total, item_discount, item_tax

  • Inventory

  • Either a v_inventory_on_hand view or an inventory table with on_hand or quantity_on_hand,
    plus a relation to products

  • Stock movements (optional)

  • stock_movements table, used by the seeder when present

The code attempts to pick the right columns dynamically and raises clear runtime errors if required tables or columns are missing.


Safety and best practices

  • Prefer analytics/reporting tools over raw SQL (revenue_by_day, sales_report, ops_health_report, etc.).
  • Lock down writes in production: set ALLOW_WRITES=false.
  • Refresh schema cache after migrations: call refresh_schema_cache.

Development

  • Code style: ruff (see [tool.ruff] in pyproject.toml)
  • Editable install:bash uv pip install -e .

You can then run ecom-mcp from your environment while iterating on the code.


License

MIT License

Copyright (c) 2025 Masoud Ahangary

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

Contributing

Issues and PRs are welcome. If you add new tools, please include:

  • expected schema assumptions
  • sample outputs
  • safety considerations (read/write permissions, timeouts, limits)

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
ecommerce-analytical-mcp-server 怎么安装?

在客户端按下面的安装命令接入:uv --directory E:\projects\ecom-mcp run ecom-mcp。装完在客户端配置里引用同一个包或端点。

ecommerce-analytical-mcp-server 怎么在 Claude Desktop、Claude、Cursor 里配置?

它的传输方式是stdio · 本地进程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

ecommerce-analytical-mcp-server 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

ecommerce-analytical-mcp-server 需要 API Key 吗?

资料没有标注鉴权方式,是否需要密钥取决于服务本身,接入前按官方文档确认。

ecommerce-analytical-mcp-server 支持哪些客户端?

资料标注支持 Claude Desktop、Claude、Cursor、VS Code、Zed;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

ecommerce-analytical-mcp-server 能提供哪些能力?

资料显示能力标签包括 Local Deployment、E-commerce Analytics、PostgreSQL、Python、AI Agent、Data Analysis。具体工具清单以官方文档为准。

ecommerce-analytical-mcp-server 需要什么运行环境?

运行时为 python;属于本地进程,需要在本机装好运行环境。

ecommerce-analytical-mcp-server 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2025-12-17。