MCP Server for Spinnaker

What is MCP Server for Spinnaker?

It is a Model Context Protocol (MCP) server that allows AI models to interact with Spinnaker deployments, pipelines, and applications via a standardized interface. It is designed for developers and DevOps engineers who want to integrate AI into their CI/CD workflows.

How to use MCP Server for Spinnaker?

Install the npm package @airjesus17/mcp-server-spinnaker or add it with yarn. Configure it by instantiating the SpinnakerMCPServer class with your Spinnaker Gate URL, a list of applications, and monitored environments, then call server.listen(port). Alternatively, set the environment variables GATE_URL, MCP_PORT, and REFRESH_INTERVAL.

Key features of MCP Server for Spinnaker

• Provides three tools: get-applications, get-pipelines, trigger-pipeline
• Automatically refreshes context data every 30 seconds
• Supports TypeScript with exported type definitions
• Configurable via environment variables
• Uses the standard MCP interface for AI integration

Use cases of MCP Server for Spinnaker

• Make intelligent deployment decisions by analyzing application and pipeline state
• Proactively detect issues and autonomously remediate (e.g., create pull requests for vulnerabilities)
• Continuously optimize CI/CD processes by identifying bottlenecks and experimenting with configurations
• Automate root cause analysis and recovery, such as rolling back to a previous version
• Act as a virtual assistant for product engineers throughout the deployment pipeline

FAQ from MCP Server for Spinnaker

What are the runtime requirements?

Node.js is required. The server communicates with a Spinnaker Gate service, so a reachable Gate URL must be provided.

Where does the context data live?

The server fetches data from Spinnaker Gate and maintains it in memory, refreshing it every 30 seconds by default.

What transport does the server use?

It listens on an HTTP port (default 3000) as configured by the MCP_PORT environment variable.

Is authentication supported?

The README does not mention any authentication mechanism; only the Gate URL and monitored applications/environments are required for configuration.