wrediam/coolify-mcp-server
Project Overview
Coolify MCP Server
An MCP server that exposes the full Coolify v4.1.1 REST API as tools for AI assistants.
True Intent (AI Assessment)
Provide AI assistants with programmatic control over a Coolify self-hosting platform via the Model Context Protocol.
README Accuracy
Notes
The test-tools.js script is misleading — it prints success messages without actually testing anything. Otherwise, stated purpose matches observed behavior and dependencies.
This is a Model Context Protocol (MCP) server written in TypeScript that acts as a bridge between MCP-compatible AI clients (like Claude Desktop, Windsurf, or Cline) and a self-hosted Coolify instance. It wraps Coolify's v4.1.1 REST API into a comprehensive set of callable tools, allowing an AI to manage infrastructure and applications through natural-language requests.
The problem it solves is operational friction. Coolify is a self-hosted PaaS (an open-source Heroku/Vercel alternative), and administering it normally requires clicking through a dashboard or hand-crafting API calls. This server lets users delegate those tasks — deploying apps, provisioning databases, restarting services, managing environment variables, spinning up Hetzner servers — to an AI agent via conversational commands.
The target users are DevOps engineers, indie developers, and homelab operators who already run Coolify and use AI coding assistants. They install the package via npm/npx, plug two environment variables (base URL and API token) into their MCP client config, and gain conversational control over their Coolify instance.
Architecturally, it's simple: a single TypeScript entry point (src/index.ts) uses the @modelcontextprotocol/sdk to register tool handlers and axios to make authenticated HTTP calls to Coolify's /api/v1 endpoints. It communicates with the MCP client over stdio. Coverage is broad — teams, servers, projects, environments, applications (public, private GitHub, deploy key, Dockerfile, Docker image, Docker Compose variants), databases (Postgres, MySQL, MariaDB, MongoDB, Redis, KeyDB, Dragonfly, ClickHouse), services, deployments, private keys, GitHub apps, cloud tokens, and Hetzner provisioning.
Notable observations: the codebase is essentially one large source file plus a smoke-test script (test-tools.js) that does not actually invoke tools — it just prints checkmarks, so 'test' claims should be taken with a grain of salt. Dependencies are minimal and reputable (only axios and the official MCP SDK). No obfuscation, no network calls beyond the user-configured Coolify host, no telemetry visible.
| Languages | TypeScript, JavaScript |
| Runtime | Node.js 18+ |
| Framework | Model Context Protocol SDK |
| Database | None (client for external Coolify API) |
| Package Manager | npm |
| Key Dependencies | @modelcontextprotocol/sdk, axios |
| Build Tool | TypeScript compiler (tsc) |
| Test Framework | None (test-tools.js is a stub script) |