port

Port MCP Server provides an interface to interact with Port's capabilities via the MCP (Model Context Protocol) framework. This server exposes tools for querying entity details, on-call status, catalog insights, creating and managing scorecards and rules, and configuring RBAC permissions. It can be consumed by clients such as Claude Desktop or Cursor, enabling AI-assisted automation and natural language interactions with your Port-backed resources. The Docker-based configuration allows you to run the server locally or in CI, exposing the Port credentials and region through environment variables so that the MCP server can authenticate with Port's services.

To use the server, configure your MCP client (Claude or Cursor) to point at the Port MCP server instance. Provide the required environment variables (PORT_CLIENT_ID, PORT_CLIENT_SECRET, PORT_REGION) and (optionally) PORT_LOG_LEVEL to adjust logging. The server will then respond to MCP actions by interfacing with Port’s APIs to fetch information, run analyses, or create resources such as scorecards or rules as specified by your queries or automation flows.

Prerequisites:

• Docker installed on your system, or a Python UVX-based installation if you prefer uvx (Port also provides a Python-based option).
• Port account with Client ID and Client Secret, and a Port region (EU or US).
• Access to the Port MCP Server image: ghcr.io/port-labs/port-mcp-server:latest.

Installation steps (Docker):

1. Ensure Docker is running on your machine.
2. Pull the Port MCP Server image (optional if you want to run directly): docker pull ghcr.io/port-labs/port-mcp-server:latest
3. Run the MCP server container with required credentials: docker run -i --rm
   -e PORT_CLIENT_ID="<PORT_CLIENT_ID>"
   -e PORT_CLIENT_SECRET="<PORT_CLIENT_SECRET>"
   -e PORT_REGION="EU"
   -e PORT_LOG_LEVEL="DEBUG"
   ghcr.io/port-labs/port-mcp-server:latest

Installation steps (UVX / Python, if preferred):

1. Install uvx and Python as per Port’s guidance.
2. Install the Port MCP server package: uvx mcp-server-port --client-id <PORT_CLIENT_ID> --client-secret <PORT_CLIENT_SECRET> --region EU --log-level DEBUG
3. Ensure environment variables PORT_CLIENT_ID, PORT_CLIENT_SECRET, PORT_REGION are set accordingly.

Prerequisites recap:

• A Port account (Sign up at Port.io).
• Client credentials (Client ID and Client Secret).
• Region (EU or US).
• Docker or uvx installation as preferred.

Tips and common considerations:

• Always provide PORT_REGION (EU or US); if omitted, EU may be used by default in some setups.
• Use PORT_LOG_LEVEL to adjust verbosity during troubleshooting (e.g., DEBUG for development).
• When using Docker, remember to mount or pass in sensitive credentials securely and avoid committing them to version control.
• If you switch from Docker to uvx, ensure the equivalent environment variables (PORT_CLIENT_ID, PORT_CLIENT_SECRET, PORT_REGION) are set, and the mcp-server-port package version is compatible with your environment.
• Claude Desktop and Cursor integration expect a valid MCP server endpoint; verify that the server is reachable and that credentials are correct to avoid authentication errors.