[!TIP] Mollie launched an official MCP Server! Go to https://docs.mollie.com/docs/mollie-mcp-server to find out more. If you want to tinker with MCP and your Mollie account, the offical MCP server is almost certainly the product you want to use.

Mollie MCP Server Proof-Of-Concept

This project implements a Model Context Protocol (MCP) server that interfaces with the Mollie API to provide information about Mollie payments.

What Is This Good For?

Since Mollie expects customers to be available for a redirect, it doesn't really make sense to do payments over an LLM conversation (except for payment links with a long runtime). So we're concentrating on information retrieval here. Users can ask for the latest transactions on their mollie account and the LLM will receive all available info on them. Based on this, we could let the LLM do creative things like advanced BI or writing Haikus about the names of your customers.

You can also get your current profile information (like profile ID, website, business email or your business category) and get a list of all supported payment methods in your account.

Additionally, you can ask the LLM to create payment links with a specific amount.

The benefit of piping your payment data to an LLM is that you can now query the data in natural language without rigid UIs.

About MCP Servers

Model Context Protocol (MCP) servers act as bridges between language models and various data sources or tools. They expose capabilities (like fetching data or executing actions) that a language model can then utilize to fulfill user requests. This allows language models to interact with external systems and access real-time information or perform tasks beyond their inherent knowledge. This server, for example, allows a model to query Mollie payment information.

Technology Used

• Node.js: Runtime environment for executing JavaScript/TypeScript.
• TypeScript: Superset of JavaScript that adds static typing.
• Mollie API Client: Official Node.js library for interacting with the Mollie API.
• dotenv: For loading environment variables from a .env file.
• @modelcontextprotocol/sdk: For building the MCP server.

Setup Instructions

1. Clone the repository:

   git clone git@github.com:hreinberger/mollie-mcp.git
   cd mollie-mcp
   

   Copy

2. Install dependencies:

   npm install
   

   Copy

3. Set up environment variables:

   • copy the example to a .env file in the root of the project:

     cp .env.example .env
     nano .env
     

     Copy
   • Replace your_mollie_api_test_or_live_key with your actual Mollie API key. You can find your API keys in your Mollie Dashboard.

4. Build the project:

   npm run build
   

   Copy

   This will compile the TypeScript files from src/ to JavaScript in the build/ directory.

5. Run the MCP Server: If you are using an MCP-compatible client (I'm using Github Copilot), you can add the config in .vscode/mcp.json.

   Example mcp.json:

   {
     "servers": {
       "mollie-mcp": {
         "type": "stdio",
         "command": "node",
         "args": ["<absolute-path-to-repo>/build/index.js"]
       }
     }
   }
   

   Copy

Alternative: HTTP + OAuth (no API key, browser-based login)

src/httpServer.ts runs this server over Streamable HTTP instead of stdio, as an OAuth-protected resource that delegates authorization to Mollie's own authorization server (https://my.mollie.com -- the same one behind Mollie's official mcp.mollie.com). No MOLLIE_API_KEY is needed for this mode: each request carries its own bearer token, obtained by the MCP client via browser-based login (PKCE, with Dynamic Client Registration where supported).

npm run build
npm run start:http   # listens on :3001 by default

See examples/openwebui-oauth for a full local walkthrough connecting this to Open WebUI's native MCP support (v0.6.31+), including why this doesn't require hosting the server publicly.

Running with Docker

Two images are published to ghcr.io/hreinberger/mollie-mcp, one per transport above -- there is no :latest tag, pick the one that matches how you want to authenticate:

• :http -- the OAuth server (src/httpServer.ts), no API key needed. See examples/openwebui-oauth for a full Open WebUI walkthrough.
• :mcpo -- the stdio server (src/index.ts) fronted by mcpo as an OpenAPI HTTP API, using a static MOLLIE_API_KEY. See examples/openwebui for a full walkthrough.

Available Tools

The server exposes the following tools that can be called by an MCP client:

• fetch-transactions:

  • Description: Get a list of the last transactions from Mollie.
  • Input: limit (number, optional, default: 5) - The number of transactions to retrieve.
  • Output: A JSON string containing the list of payment objects from Mollie.

• create-payment-link:

  • Description: Create a payment link with a specified amount.
  • Input: amount (string) - The amount for the payment, formatted as a string with a dot for the decimal separator (e.g., "123.45").
  • Output: A string containing the direct URL to the Mollie payment link.

• transaction-info:

  • Description: Get all info for a given payment transaction. The transaction ID must start with "tr_".
  • Input: paymentId (string) - The Mollie payment transaction ID (e.g., "tr_xxx").
  • Output: A JSON string containing the full payment object for the specified transaction.

• fetch-methods:

  • Description: Get a list of all available payment methods from Mollie.
  • Input: None.
  • Output: A JSON string containing the list of available payment methods from Mollie.

• profile-info:

  • Description: Get information about the currently used Mollie profile.
  • Input: None.
  • Output: A JSON string containing the profile information for the current Mollie account.

This MCP server is designed to be a simple example. It can be extended with more tools to interact further with the Mollie API or other services.

Caveats

Currently, we're handing back the full payment objects to the LLM, which include lots of data and thus, tokens. Be mindful of your LLM's context window.