shopify-mcp-server
Exposes Shopify Admin API as typed MCP tools for inventory, orders, revenue, and more, enabling Discord-based store management via AI agents.
README
Shopify Discord Agent
Ask your Shopify store questions in Discord. Type what you need — stock levels, today's revenue, order status, discount codes — and get live answers without opening the admin.
Built with Discord, n8n, Google Gemini (Vertex AI), and a custom MCP tool server that wraps the Shopify Admin GraphQL API.
Why I built this
- Problem — E-commerce ops teams live in Discord, but every stock check or revenue question meant opening Shopify admin.
- Approach — Expose Shopify as typed MCP tools (not brittle n8n HTTP nodes) and let Gemini pick the right action via an n8n AI Agent.
- Outcome — 28 tools, mock mode for zero-credential demos, and a CONFIRM gate on high-risk writes so the agent can act safely in production chat.
<p align="center"> <img src="assets/images/hero-workspace.png" alt="Shopify Discord Agent — workspace hero" width="100%" /> </p>
<p align="center"> <a href="#quick-start-docker"><strong>Quick start</strong></a> · <a href="#demo-video">Demo</a> · <a href="#what-it-does">Features</a> · <a href="#architecture">Architecture</a> · <a href="docs/n8n-setup.md">n8n setup</a> · <a href="docs/docker-setup.md">Docker guide</a> </p>
Demo video
https://github.com/Ai-assistant-KIRA/shopify-discord-agent/assets/main/assets/demo.mp4
Discord → n8n → Gemini → MCP → Shopify Admin API in one conversational flow.
Quick start (Docker)
One-time setup — credentials are saved to config/.env and loaded on every start (permanent across restarts):
git clone https://github.com/Ai-assistant-KIRA/shopify-discord-agent.git
cd shopify-discord-agent
npm install
npm run setup # once: Shopify + Discord credentials → config/.env
npm run docker:up # starts stack; auto-restarts on reboot
Then (first launch only):
- Open http://localhost:5678 and create your n8n account
- Import
n8n-workflows/discord-shopify-mcp-agent.docker.json - Add Google Vertex credentials and set your GCP project ID in the Chat Model node
- Activate the workflow
- Message your bot in Discord
Vertex credentials are also saved once — they persist in the n8n Docker volume.
Guides: n8n workflow setup (local + cloud) · Docker walkthrough · Shopify credentials
Mock demo (no Shopify store)
npm run setup # choose mock mode when prompted
npm run docker:mock
Enable write tools later
Edit config/.env and set SHOPIFY_READ_ONLY=false, then docker compose restart mcp.
What it does
Your team stays in Discord. The agent handles the Shopify lookups and actions.
| Category | Examples |
|---|---|
| Inventory | Stock by SKU, low-stock alerts, quantity updates |
| Orders | Search, daily list, fulfillment status, cancel, refund |
| Revenue | Today, yesterday, last 7/30 days |
| Products | Search catalog, list variants, pricing, publish/draft |
| Discounts | Create codes, list active promos, deactivate |
| Draft orders | Create, list, send invoices |
| Customers | Search, order history |
| Carts | Abandoned checkout summary |
28 MCP tools in full write mode. Read-only mode hides write tools for safe demos.
<p align="center"> <img src="assets/images/architecture-flow.png" alt="Architecture — Discord, n8n, Gemini, MCP, Shopify" width="90%" /> </p>
Example prompts
What's the stock for BB-TONER-200?
How much revenue did we make today?
Show unfulfilled orders from this week
Create a 15% code MUGWORT15 for the toner collection
Draft order for jane@example.com — 2x BB-TONER-200
Which products are below 10 units?
<p align="center"> <img src="assets/images/discord-chat.png" alt="Discord chat interface for store operations" width="90%" /> </p>
Architecture
flowchart LR
User[Discord user] --> Bridge[discord-bridge]
Bridge --> Webhook[n8n webhook]
Webhook --> Agent[AI Agent + Gemini]
Agent --> MCP[MCP Client Tool]
MCP --> Server[shopify-mcp-server]
Server --> Shopify[Shopify Admin API]
Agent --> Reply[Webhook response]
Reply --> Bridge
Bridge --> User
| Layer | Role |
|---|---|
| Discord bridge | Forwards channel messages to n8n; posts replies back |
| n8n workflow | LangChain AI Agent with per-channel memory |
| Gemini (Vertex) | Understands intent, picks tools, formats answers |
| MCP server | 28 typed tools over SSE — inventory, orders, discounts, etc. |
| Shopify | Live Admin GraphQL (or mock data for demos) |
Details: docs/architecture.md
Safety
High-risk writes (inventory changes, fulfillments, refunds, price updates) return a CONFIRM token. The agent waits for explicit approval before executing.
SHOPIFY_READ_ONLY=true— hides all write toolsSHOPIFY_MOCK_MODE=true— sample data, zero Shopify credentials needed
Production use
After npm run setup, users still need a one-time n8n workflow import (Vertex credential — persists in Docker volume). Then:
npm run docker:up
npm run verify:production # pre-flight check
| Ready for | Requirements |
|---|---|
| Read-only pilot | config/.env + active n8n workflow + channel allowlist |
| Daily ops | Above + HTTPS if remote + MCP_API_KEY if MCP is public |
| Write-enabled | Above + SHOPIFY_READ_ONLY=false + team CONFIRM training |
Full checklist: docs/production-checklist.md
Project layout
shopify-discord-agent/
├── config/ # Permanent credentials (config/.env — created by npm run setup)
├── assets/ # Demo video + README images
├── docs/ # Setup guides (incl. n8n local + cloud)
├── lib/ # Shopify client, tool registry, mock data
├── n8n-workflows/ # Importable workflow JSON
├── scripts/
│ ├── discord-bridge.mjs # Discord ↔ n8n
│ ├── import-workflow.mjs
│ └── verify-shopify-creds.mjs
├── tests/ # MCP integration tests
├── tools/ # MCP tool modules (28 tools)
├── docker-compose.yml
├── shopify-mcp-server.mjs
└── README.md
Local development (no Docker)
npm install
cp .env.example .env
npm run mcp:mock # terminal 1 — MCP on :3000
npx n8n start # terminal 2 — n8n on :5678
Import n8n-workflows/discord-shopify-mcp-agent.json, configure Vertex, activate, then:
npm run bridge # terminal 3 — Discord listener
Health check: http://localhost:3000/health
For n8n Cloud, the MCP server must be on a public HTTPS URL — see docs/n8n-setup.md.
Setup guides
- n8n workflow — local self-hosted and n8n Cloud
- Docker Compose — recommended one-command stack
- Discord bot — token, intents, channel setup
- Customer demo server —
npm run discord:demo-setup - Shopify credentials — dev store + Admin API token
- Architecture — tools, safety tiers, data flow
- Production checklist — go-live hardening
Environment variables
| Variable | Default | Purpose |
|---|---|---|
SHOPIFY_MOCK_MODE |
true |
Sample data instead of live API |
SHOPIFY_READ_ONLY |
true |
Hide write tools |
SHOPIFY_DOMAIN |
— | Shopify store domain |
SHOPIFY_ACCESS_TOKEN |
— | Admin API token (shpat_…) |
DISCORD_BOT_TOKEN |
— | Bot token for the bridge |
DISCORD_ALLOWED_CHANNEL_IDS |
— | Comma-separated channel IDs (recommended) |
MCP_API_KEY |
auto-generated | Bearer auth on MCP SSE (required if public) |
N8N_WEBHOOK_URL |
http://localhost:5678/webhook/discord |
Production webhook |
PORT |
3000 |
MCP server port |
npm scripts
| Command | What it does |
|---|---|
npm run setup |
One-time credential wizard → config/.env |
npm run docker:mock |
Start full stack with mock Shopify |
npm run docker:up |
Start stack with live Shopify |
npm run mcp:mock |
MCP server only (mock) |
npm run bridge |
Discord → n8n bridge |
npm run verify:shopify |
Test API credentials |
npm run verify:production |
Pre-flight production check |
npm run discord:demo-setup |
Create demo Discord channels |
npm test |
MCP integration tests |
Prerequisites
- Node.js 20+
- Docker (for the one-command setup)
- Discord bot with Message Content intent
- Google Cloud project with Vertex AI (Gemini)
- Shopify dev store (optional — mock mode works without it)
Troubleshooting
| Issue | Fix |
|---|---|
| No Discord reply | Workflow must be Active; check npm run bridge logs |
| MCP errors | curl localhost:3000/health — start MCP with npm run mcp:mock |
| Shopify 401 | Use Admin API token (shpat_), not session token |
| Webhook 404 | Activate workflow or check webhook path |
| Vertex errors | Set GCP project ID in the Chat Model node |
License
MIT — see LICENSE.
<p align="center"> <strong>Built by <a href="https://www.linkedin.com/in/reda-alaarabi">Reda Alaarabi</a></strong><br> <sub>Automation · e-commerce integrations · AI agents</sub> </p>
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。