Overview

The Adobe Commerce MCP Server is a lightweight proxy service that implements the Model Context Protocol (MCP) to simplify integrations with Adobe Commerce (Magento) via its GraphQL API. It exposes a small set of RESTful endpoints for working with orders, products, and customers, translating requests into GraphQL operations against your Adobe Commerce instance. This reduces the amount of GraphQL boilerplate in client apps, automations, and AI agents that need to read or mutate store data.

This server is useful when you want a predictable, authentication-friendly HTTP interface in front of Adobe Commerce, for local development, automation, or to satisfy integration patterns where GraphQL clients are not ideal. It also centralizes credentials and request shaping, so you can apply consistent validation, logging, and rate-limiting in one place.

Features

REST-style endpoints for Orders, Products, and Customers mapped to Adobe Commerce GraphQL
Lightweight Express-based server suitable for local development and cloud deployments
Centralized configuration for Adobe Commerce GraphQL endpoint and access token
Simple JSON request/response model for easier integration with automation tools and agents
Optional webhook signing/verification and logging hooks for observability
Docker-friendly for quick local or CI usage

Installation / Configuration

Clone and install:

git clone https://github.com/rafaelstz/adobe-commerce-dev-mcp.git
cd adobe-commerce-dev-mcp
npm install

Environment variables (create a .env from .env.example):

# .env
PORT=3000
COMMERCE_GRAPHQL_URL=https://your-magento-instance.com/graphql
COMMERCE_ACCESS_TOKEN=your-integration-access-token
LOG_LEVEL=info
WEBHOOK_SECRET=optional-secret-for-signing

Start the server locally:

# development
npm run dev

# production
npm start

Run with Docker (example Dockerfile or image provided in the repo):

docker build -t adobe-commerce-mcp .
docker run -p 3000:3000 \
  -e COMMERCE_GRAPHQL_URL=https://your-magento-instance.com/graphql \
  -e COMMERCE_ACCESS_TOKEN=your-access-token \
  adobe-commerce-mcp

Available Resources

GitHub repository: https://github.com/rafaelstz/adobe-commerce-dev-mcp
Adobe Commerce (Magento) GraphQL docs: https://developer.adobe.com/commerce/graphql/
Example .env and Dockerfiles included in the repository
Built-in request logging and simple error handling for debugging

API Endpoints

Below is a concise overview of commonly available endpoints. Exact paths may vary slightly by release; consult the repository for the most current routes.

Method	Path	Purpose
GET	/api/mcp/products	Search or fetch product listings (query params: search, sku, limit, offset)
GET	/api/mcp/products/:sku	Fetch product details by SKU
POST	/api/mcp/customers	Create a customer (JSON body: customer data)
GET	/api/mcp/customers/:id	Fetch customer by ID
GET	/api/mcp/orders	List orders (filters via query params)
GET	/api/mcp/orders/:id	Fetch order details by ID
POST	/api/mcp/orders	Create or update orders (depending on payload and store support)

Example: fetch products via curl

curl "http://localhost:3000/api/mcp/products?search=shirt&limit=10"

Create a customer:

curl -X POST "http://localhost:3000/api/mcp/customers" \
  -H "Content-Type: application/json" \
  -d '{
    "email": "[email protected]",
    "firstname": "Jane",
    "lastname": "Doe",
    "password": "safepass123"
  }'

The MCP server translates these requests into the corresponding GraphQL operations and returns normalized JSON responses.

Use Cases

Automations and agents: Use the MCP server as a stable HTTP interface for automation scripts, RPA tools, or LLM agents that need to query or mutate store data without embedding GraphQL logic.
Local development: Spin up a local proxy that points to a staging Adobe Commerce instance so frontend developers can work with a simplified REST surface while the store evolves.
ETL and reporting: Periodically fetch product or order data through predictable endpoints, then push into analytics pipelines or data warehouses without constructing GraphQL queries each time.
Webhook processing: Centralize inbound webhook verification and then use MCP endpoints to enrich or mutate order/customer state in Adobe Commerce.
Microservice integrations: Other backend services can call the MCP server to keep Adobe Commerce interactions encapsulated (token management, retries, and logging handled in one place).

Tips for Developers

Keep your COMMERCE_ACCESS_TOKEN secret and scope it to only the permissions needed for the operations your MCP instance performs.
For high-throughput use, place the MCP server behind a caching layer (e.g., Varnish, Redis) for product read endpoints and implement rate limiting to protect the Adobe Commerce GraphQL endpoint.
Extend the server to add custom GraphQL queries or mutations exposed via additional REST endpoints if you have domain-specific workflows (subscriptions, custom attributes, etc.).
Run the included tests and inspect logs to understand request/response mapping during onboarding.

For implementation details, example queries, and the latest code, see the project repository: https://github.com/rafaelstz/adobe-commerce-dev-mcp.

Adobe Commerce MCP Server: Orders, Products, Customers