Best for
- Developers building e-commerce on Medusa JS who want AI-agent automation
- Teams integrating a Medusa backend with LLM toolchains
- Engineers comfortable customizing OpenAPI definitions for fine-grained tool control
M
43FMRS · D
A community MCP server for the Medusa JS SDK focused on extensibility and AI automation, supporting OpenAPI-based customization and conversion into a Claude Skill. Best suited to developers with Medusa and Node.js experience, but it lacks official backing and detailed tool documentation, so security and functionality should be assessed before use.
medusa-mcp is a Model Context Protocol (MCP) server designed for integration with the Medusa JavaScript SDK. It provides a scalable backend layer for managing and interacting with Medusa's data models, enabling automation, orchestration, and intelligent service extensions. It can automate workflows such as inventory or pricing adjustments, connect external tools like email and analytics, and let AI agents analyze trends and trigger actions. It is an MIT-licensed community project, not maintained by Medusa officially.
Setup
- Clone https://github.com/SGFGOV/medusa-mcp and run npm install; 2. Build with npm run build; 3. Configure environment variables: MEDUSA_BACKEND_URL (Medusa backend URL), PUBLISHABLE_KEY (publishable API key), MEDUSA_USERNAME and MEDUSA_PASSWORD (admin credentials); 4. Start the server with npm start; 5. Optionally test with npx @modelcontextprotocol/inspector ./dist/index.js (restart the Inspector and browser after each rebuild).
Fit and risk
Not for
- Projects not using Medusa
- Users seeking official Medusa support or hosted solutions
- Non-technical users unable to deploy a Node.js service and manage credential environment variables
Required permissions
- Network access to a Medusa backend URL
- A Medusa publishable API key (PUBLISHABLE_KEY)
- Medusa admin username and password for admin operations
Risks and side effects
- Admin credentials stored as environment variables carry leak risk
- AI agents can trigger real commerce operations (order fulfillment, pricing updates); mistakes may affect production data
- Server defaults to localhost:3000; exposing it to a network requires your own security measures
- No tool list is documented, so actual capabilities must be verified by reviewing the source
Troubleshooting
- Restart the MCP Inspector and browser after every rebuild
- Verify MEDUSA_BACKEND_URL and PUBLISHABLE_KEY are set correctly
- If admin tools fail, confirm MEDUSA_USERNAME and MEDUSA_PASSWORD are provided
- On build errors, run npm install first to ensure dependencies are complete
- For customization, check that oas/admin. and oas/store. contain up-to-date OAS definitions
Use cases
Automating commerce workflows such as inventory or pricing adjustments
Letting AI agents access product and order data from a Medusa backend
Customizing admin and store tools via OpenAPI definitions (oas/admin., oas/store.)
Converting the server into a Claude Skill using the included mcp-to-skill.py script