cpi

The CPI MCP server exposes a suite of AI-assisted tools for SAP Cloud Integration, including retrieval and search operations for integration packages, integration flows, value mappings, message mappings, script collections, and integration runtime artifacts. The server is designed to be run in a container (Docker) or as a standalone binary, and it surfaces endpoints that correspond to the listed tools (e.g., get_integration_packages, search_integration_packages, get_integration_flows, etc.). To use the server, ensure your environment variables are configured with your SAP Cloud Integration tenant and credentials, then start the container or binary. Once running, you can invoke the available tools via HTTP/REST requests to the server, passing parameters such as IDs, names, versions, or statuses to filter results. For example, you can fetch all integration packages or search for specific flows by name or ID, enabling AI-assisted discovery and mapping against your SAP Integration Suite assets. The server supports both standard retrieval and search operations, making it convenient to programmatically inventory and locate artifacts within your tenant. When using HTTP transport, you can target the host and port exposed by the container or binary and execute the corresponding tool endpoints as defined by the MCP server's API surface.

Prerequisites:

• Docker installed and running (recommended) or a compatible Go toolchain if you build from source.
• Access credentials and endpoints for your SAP Cloud Integration (MCP) account (Base URL, Token URL, Client ID, Client Secret).

Installation steps (Docker):

1. Pull the MCP server Docker image:

docker pull vadimklimov/cpi-mcp-server

2. Run the container with required environment variables (example values shown; replace with your own):

docker run -d \
  --name cpi-mcp \
  -e MCP_CPI_BASE_URL=https://your-cpi.example.com \
  -e MCP_CPI_TOKEN_URL=https://your-cpi.example.com/token \
  -e MCP_CPI_CLIENT_ID=your-client-id \
  -e MCP_CPI_CLIENT_SECRET=your-client-secret \
  -e MCP_CPI_TRANSPORT=http \
  -e MCP_CPI_PORT=8080 \
  vadimklimov/cpi-mcp-server

3. Verify the container is running:

docker ps | grep cpi-mcp

If you prefer building from source or using a standalone binary, follow the repository's Go-based build steps in the Docker-less path below:

Installation steps (Standalone binary):

1. Build from source (if you have Go installed):

git clone https://github.com/vadimklimov/cpi-mcp-server.git
cd cpi-mcp-server
go build -o cpi-mcp-server .

2. Run the binary with required environment variables or a config file (example):

export MCP_CPI_BASE_URL=https://your-cpi.example.com
export MCP_CPI_TOKEN_URL=https://your-cpi.example.com/token
export MCP_CPI_CLIENT_ID=your-client-id
export MCP_CPI_CLIENT_SECRET=your-client-secret
export MCP_CPI_TRANSPORT=http
export MCP_CPI_PORT=8080
./cpi-mcp-server

3. Confirm the server is listening on the specified port and ready to accept tool requests.

Tips and common considerations: - Always provide the required CPI base URL, token URL, client id, and client secret; without these, authentication may fail. - If you plan to expose the API over HTTP, ensure proper security measures (TLS termination, credentials management). - The transport option can be stdio or http; use http for easier programmatic access. - The MCP_CPI_PORT defaults to 8080 when not specified; ensure the port is open in your container or host firewall. - When using Docker, you can map ports to access the API from outside the container, e.g., -p 8080:8080. - Review logging options (MCP_CPI_LOG_LEVEL, MCP_CPI_LOG_FILE) to tailor verbosity and log retention for production environments.