Commerce MCP Server

commerce-mcp-server

Introduction

This repository contains an MCP (Model Context Protocol) server that exposes tools for working with Commerce OData APIs.

It supports two common ways of running:

• Local development: Run the server locally and use the MCP Inspector to explore and call tools.
• SAP BTP (Cloud Foundry): Build and deploy as an MTA, using BTP services (e.g., Destination and XSUAA) for connectivity and authentication.

I highly recommend to go through the complete README file and then start step by step

Pre-requisites

• SAP BTP subaccount (Cloud Foundry): You have access to a BTP subaccount with a Cloud Foundry org/space and can log in using cf login.

  Not required if running locally with MOCK_BACKEND=true.

• Commerce OData APIs available:
  1. Commerce OData APIs are already deployed/running and reachable from where you run this MCP server (local or BTP). You can either create a simple e-commerce application using standard CAP template or create your own.
  2. Create a destination for the commerce APIs with name - COMMERCE_API_DESTINATION.

  Not required if running locally with MOCK_BACKEND=true.

Destination name default:

• Default Destination name is COMMERCE_API_DESTINATION which can be updated from envrironment variables

Local Set up

1. Install Python 3.13 and verify

Install Python 3.13.x from python.org (macOS universal installer).

Verify:

python3.13 --version

python3.13 --version
which python3.13

1. (Optional) Open new terminal and install uv (faster pip) and ensure correct architecture (arm64 vs x86_64)

curl -LsSf https://astral.sh/uv/install.sh | sh

1. (Optional) Add uv to your PATH

The install script typically places uv in ~/.local/bin/uv. Adding ~/.local/bin to your PATH makes uv available in new terminals.

If you installed uv via Homebrew (brew install uv), you can skip this step.

echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.zshrc
source ~/.zshrc
hash -r

1. (Optional) Verify uv is found

which uv
uv --version
file "$(which uv)"

1. Go to VSCode and open terminal. Create and activate a virtual environment

python3.13 -m venv .venv
source .venv/bin/activate

Verify venv is active and using expected Python:

which python
python --version
python -m pip --version

1. Install fastMCP (uv optional)

Using uv:

uv pip install --no-cache-dir "fastmcp<3"

Or using pip:

pip install "fastmcp<3"

1. Install dependencies

pip install -r requirements.txt

1. Verify version

fastmcp version

You are all set!

MCP Server on BTP and Quick Validation

1. Login to the BTP subaccount using cf login

2. Build application and deploy to BTP

mbt build && cf deploy mta_archives/commerce-mcp-server_1.0.0.mtar

1. The MBT build will create the required resources and deploy the MCP server in the subaccount dev space.

2. Run the below MCP client which will list all the tools of the MCP. Ensure you have your .env set up properly as per the .env-template

python scripts/check_remote_mcp.py

1. You will receive the list of tools

Connected.
Tools: ['mcp_ping']
mcp_ping: {'status': 'success', 'message': 'MCP service is reachable.'}

VCAP SERVICES

1. Get the vcap services

cf env commerce-mcp-server

1. Create vcap.json at the root of your folder. Copy the full JSON under VCAP_SERVICES and save it as vcap.json.

2. Set env variables

export VCAP_SERVICES="$(cat vcap.json)"

1. Check if the VCAP_SERVICES are loaded using below command

echo $VCAP_SERVICES

MCP server on local

There are two ways to run the MCP server locally:

Option 1: Mock Backend (no BTP required)

No SAP BTP account or deployed Commerce APIs needed. The server serves sample payloads from assets/payloads/.

1. Create a .env file in the project root using the .env-template and set:

ENVIRONMENT=local
MOCK_BACKEND=true

1. Run MCP inspector (local STDIO)

fastmcp dev src/server.py

1. This opens the MCP Inspector in your browser. All tools (get_products, get_categories, get_products_by_category, get_retailers) return mock data immediately.

Option 2: Live Backend (BTP required)

Pre-requisite - Commerce OData APIs application must be deployed on BTP with DESTINATION - COMMERCE_API_DESTINATION.

1. Create a .env file in the project root using the .env-template and set:

ENVIRONMENT=local
MOCK_BACKEND=false
DESTINATION_NAME=COMMERCE_API_DESTINATION

1. Follow the VCAP SERVICES section to load your service credentials.

2. Run MCP inspector (local STDIO)

fastmcp dev src/server.py

1. This opens the MCP Inspector in your browser.

2. Browse through the tools and run them.

How to debug the code on local

1. Start the Debug launch configuration.

2. In Terminal 1 (debug console/terminal), verify the server is running and the MCP endpoint is available: http://0.0.0.0:8080/mcp

3. Open second terminal and load the xsuaa service credentials (You will get it from subaccount xsuaa service instance)

export MCP_URL=http://0.0.0.0:8080/mcp
export MCP_XSUAA_AUTH_URL=https://<subaccount>.authentication.us10.hana.ondemand.com/oauth/token
export MCP_XSUAA_CLIENT_ID=<clientID>
export MCP_XSUAA_CLIENT_SECRET=<clientSecret>

1. Get MCP_XSUAA_TOKEN (XSUAA client credentials)

If you see KeyError: 'access_token', the token endpoint returned an error JSON (so there is no access_token field). First, inspect the raw response:

curl -i -sS -X POST "$MCP_XSUAA_AUTH_URL" \
  -u "$MCP_XSUAA_CLIENT_ID:$MCP_XSUAA_CLIENT_SECRET" \
  -H "Accept: application/json" \
  -d "grant_type=client_credentials"

Then export the token:

export MCP_XSUAA_TOKEN=$(
  curl -sS -X POST "$MCP_XSUAA_AUTH_URL" \
    -u "$MCP_XSUAA_CLIENT_ID:$MCP_XSUAA_CLIENT_SECRET" \
    -H "Accept: application/json" \
    -d "grant_type=client_credentials" \
  | python3 -c 'import json,sys; print(json.load(sys.stdin)["access_token"])'
)

1. Initialize and capture SID

SID=$(
  curl -i -sS -X POST "$MCP_URL" \
    -H "Authorization: Bearer $MCP_XSUAA_TOKEN" \
    -H "Accept: application/json, text/event-stream" \
    -H "Content-Type: application/json" \
    -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-03-26","capabilities":{},"clientInfo":{"name":"curl","version":"0.1.0"}}}' \
  | awk -F': ' 'tolower($1)=="mcp-session-id"{print $2}' | tr -d '\r'
)
echo "SID=[$SID]"

1. If SID prints, run the below command

curl -sS -X POST "$MCP_URL" \
  -H "Authorization: Bearer $MCP_XSUAA_TOKEN" \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -H "Mcp-Session-Id: $SID" \
  -d '{"jsonrpc":"2.0","method":"notifications/initialized"}'

1. List tools

curl -sS -X POST "$MCP_URL" \
  -H "Authorization: Bearer $MCP_XSUAA_TOKEN" \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -H "Mcp-Session-Id: $SID" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/list"}'

1. Set debug point in the mcp_ping tool and run the below command

curl -sS -X POST "$MCP_URL" \
  -H "Authorization: Bearer $MCP_XSUAA_TOKEN" \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -H "Mcp-Session-Id: $SID" \
  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"mcp_ping","arguments":{}}}'

1. Call get_products tool

curl -sS -X POST "$MCP_URL" \
  -H "Authorization: Bearer $MCP_XSUAA_TOKEN" \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -H "Mcp-Session-Id: $SID" \
  -d "{\"jsonrpc\":\"2.0\",\"id\":4,\"method\":\"tools/call\",\"params\":{\"name\":\"get_products\",\"arguments\":{}}}"

1. Call get_categories tool

curl -sS -X POST "$MCP_URL" \
  -H "Authorization: Bearer $MCP_XSUAA_TOKEN" \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -H "Mcp-Session-Id: $SID" \
  -d '{"jsonrpc":"2.0","id":5,"method":"tools/call","params":{"name":"get_categories","arguments":{}}}'

1. Call get_products_by_category tool

CATEGORY_ID="a1b2c3d4-e5f6-7a8b-9c0d-1e2f3a4b5c3d"

curl -sS -X POST "$MCP_URL" \
  -H "Authorization: Bearer $MCP_XSUAA_TOKEN" \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -H "Mcp-Session-Id: $SID" \
  -d "{\"jsonrpc\":\"2.0\",\"id\":6,\"method\":\"tools/call\",\"params\":{\"name\":\"get_products_by_category\",\"arguments\":{\"category_id\":\"$CATEGORY_ID\"}}}"

1. Call get_retailers tool

curl -sS -X POST "$MCP_URL" \
  -H "Authorization: Bearer $MCP_XSUAA_TOKEN" \
  -H "Accept: application/json, text/event-stream" \
  -H "Content-Type: application/json" \
  -H "Mcp-Session-Id: $SID" \
  -d '{"jsonrpc":"2.0","id":7,"method":"tools/call","params":{"name":"get_retailers","arguments":{}}}'