coolify
MCP server for Coolify — 38 optimized tools for managing self-hosted PaaS through AI assistants
claude mcp add --transport stdio stumason-coolify-mcp npx -y @masonator/coolify-mcp \ --env COOLIFY_BASE_URL="https://your-coolify-instance.com" \ --env COOLIFY_ACCESS_TOKEN="your-api-token"
How to use
This Coolify MCP server provides 38 token-optimized tools to manage and debug your Coolify installations through an AI assistant. It aggregates actions across infrastructure, diagnostics, deployments, environments, apps, databases, services, and more, returning compact, hyper-relevant summaries by default to preserve context window space. Use the high-level discovery tools like list_servers and get_infrastructure_overview to understand your setup, then drill down with get_* and resource-specific tools (e.g., get_application, deployment, database) to inspect details or perform operations. The server supports batch operations, environment variable management, and private keys, enabling robust control and automation from a single MCP endpoint.
How to install
Prerequisites:
- Node.js v18 or later installed on the host
- Access to a running Coolify instance and a valid API token
Installation steps:
- Install or run via npx (recommended for quick start):
npx -y @masonator/coolify-mcp
- If you prefer a global install (optional):
npm i -g @masonator/coolify-mcp
- Configure environment variables for the MCP to access Coolify:
COOLIFY_BASE_URL=https://your-coolify-instance.com
COOLIFY_ACCESS_TOKEN=your-api-token
- Start the MCP server (example if the package exposes a run script):
npx @masonator/coolify-mcp
Note: The README examples show integrating with Claude Desktop/Claude Code or Cursor by setting up the server under the mcpServers configuration. Ensure your environment variables are set in the environment or provided in your MCP launcher configuration when starting the service.
Additional notes
Tips and considerations:
- Make sure COOLIFY_BASE_URL uses the public URL reachable by the AI assistant. If your Coolify instance is private, set up the appropriate VPN or proxy access.
- Treat the MCP as a command surface: most tools accept an action parameter to modify resources (create, update, delete) or perform operations (restart, redeploy).
- Use get_infrastructure_overview first to get a concise snapshot of all resources, then use list_* for discovery and get_* for detailed views.
- The server optimizes responses to reduce token usage; expect compact summaries for list_* endpoints and full details only when using get_* endpoints.
- If you encounter authentication errors, verify the API token has the correct scopes and that COOLIFY_BASE_URL is correct and reachable from the deployment host.
Related MCP Servers
openapi
OpenAPI definitions, converters and LLM function calling schema composer.
CanvasMCPClient
Canvas MCP Client is an open-source, self-hostable dashboard application built around an infinite, zoomable, and pannable canvas. It provides a unified interface for interacting with multiple MCP (Model Context Protocol) servers through a flexible, widget-based system.
opnsense
Modular MCP server for OPNsense firewall management - 88 tools providing access to 2000+ methods through AI assistants
jenkins -enterprise
The most advanced Jenkins MCP server available - Enterprise debugging, multi-instance management, AI-powered failure analysis, vector search, and configurable diagnostics for complex CI/CD pipelines.
docmole
Dig through any documentation with AI - MCP server for Claude, Cursor, and other AI assistants
skill-to
Convert AI Skills (Claude Skills format) to MCP server resources - Part of BioContextAI