Coolify MCP Server
MCP server that integrates with Coolify to let AI assistants manage Coolify instances via a clean toolkit wrapping the official REST API.
README
Coolify MCP Server
A Model Context Protocol (MCP) server that integrates with Coolify to let AI assistants manage your Coolify instance via a clean toolkit that wraps the official REST API.
Table of Contents
Features
Core Capabilities
- Coverage of key Coolify API endpoints (applications, services, deployments, teams, private keys, domains, health/version)
- Team Management – List teams, get team details, and view members
- Server & Domain Insights – Query domains configured per server
- Application Lifecycle – Start/stop/restart and create applications
- Service Management – List/create/start/stop/restart services
- Deployments – List running deployments; fetch deployment by UUID
- Private Keys – List and create SSH private keys for server auth
- Env Vars – CRUD for application/service environment variables
Note: Some aggregated operations (e.g., "resources by server") are implemented by client-side filtering over the supported endpoints rather than a dedicated API path.
Prerequisites
- Node.js v18 or newer
- npm or yarn
- A running Coolify instance (self‑hosted or cloud)
- A Coolify API token with appropriate permissions
Installation
1) Clone & install
git clone https://github.com/forsonny/Coolify-MCP-Server-for-Claude-Code.git
cd Coolify-MCP-Server-for-Claude-Code
npm install
2) Build
npm run build
(Compiles TypeScript into dist/)
Configuration
Environment Setup
Create a .env file at the project root:
# Required
COOLIFY_BASE_URL=https://your-coolify-instance.com
COOLIFY_API_TOKEN=your-api-token-here
# Optional
COOLIFY_TIMEOUT=30000
Notes
- Use a full base URL (no trailing slash). Example:
https://coolify.example.com - Do not use inline comments in
.envvalues. - Never commit
.envto version control.
Claude Code Setup
You can connect the server to Claude Code using the automated setup script or manually.
Option A — Automated Setup (Recommended)
The easiest way is to use the setup script:
# Run the setup script
npm run setup
This will:
- Validate your
.envfile configuration - Build the TypeScript project if needed
- Generate the correct
claude mcp add-jsoncommand with environment variables - Provide troubleshooting tips
Option B — Manual Setup
If you prefer to set it up manually, you have two approaches:
Method 1: With explicit environment variables (most reliable)
claude mcp add-json coolify '{
"type": "stdio",
"command": "node",
"args": ["/absolute/path/to/Coolify-MCP-Server-for-Claude-Code/dist/index.js"],
"env": {
"COOLIFY_BASE_URL": "https://your-coolify-instance.com",
"COOLIFY_API_TOKEN": "your-api-token-here",
"COOLIFY_TIMEOUT": "30000"
}
}' -s local
Method 2: Simple command (relies on .env file)
claude mcp add coolify "node /absolute/path/to/Coolify-MCP-Server-for-Claude-Code/dist/index.js" -s local
Verify the connection
claude mcp list
Option C — Import from Claude Desktop If you already have the server configured in Claude Desktop:
claude mcp add-from-claude-desktop
claude mcp list
Important: Method 1 (explicit environment variables) is more reliable because it ensures environment variables are available to the MCP server process, regardless of the working directory from which Claude Code launches the server.
Tip: Some setups store Claude Code MCP config in
~/.claude.json. The Desktop app usesclaude_desktop_config.jsonunder your OS application data directory.
Cursor Setup
Cursor supports MCP. The easiest route is through Settings → Extensions → MCP Servers → Add Server, then point to your built script (node …/dist/index.js) and add the three environment variables. Restart Cursor after saving.
For advanced/enterprise setups, see Cursor's MCP docs on programmatic registration.
API Token Generation
- Open your Coolify dashboard.
- Navigate to Keys & Tokens → API tokens.
- Create a new token (name it e.g. "MCP Server"), select the permissions you need, and optionally set an expiration.
- Copy the token once when shown and paste it into your
.env.
Available Tools
System
| Tool | What it does |
|---|---|
get_version |
Get Coolify version |
health_check |
Healthcheck probe |
Teams
| Tool | What it does |
|---|---|
list_teams |
List all teams |
get_team |
Get team by ID |
get_current_team |
Get current team |
get_current_team_members |
List members of current team |
Servers & Domains
| Tool | What it does |
|---|---|
get_server_domains |
Get domains for a server UUID |
Applications
| Tool | What it does |
|---|---|
list_applications |
List applications |
create_application |
Create application |
start_application |
Start application |
stop_application |
Stop application |
restart_application |
Restart application |
Services
| Tool | What it does |
|---|---|
list_services |
List services |
create_service |
Create service |
start_service |
Start service |
stop_service |
Stop service |
restart_service |
Restart service |
Deployments & Keys
| Tool | What it does |
|---|---|
list_deployments |
List running deployments |
get_deployment |
Get deployment by UUID |
list_private_keys |
List SSH private keys |
create_private_key |
Create a new SSH private key |
Environment Variables
Application and service env vars support: list, create, update (single/bulk), and delete.
Usage Examples
Natural language you can try once connected:
Basics
"List my applications"
"Who's in the current team?"
"What version is my Coolify instance?"
Servers & Apps
"Show domains for server 123e4567-e89b-12d3-a456-426614174000"
"Create a new app from https://github.com/user/repo"
"Restart the backend app"
Services & Env Vars
"List running services"
"Create a PostgreSQL service on the main server"
"Add DATABASE_URL to the frontend app"
Development
Dev mode (watch):
npm run dev
Build:
npm run build
Setup:
npm run setup
Clean:
npm run clean
Project Structure
Coolify-MCP-Server-for-Claude-Code/
├── src/
│ ├── index.ts # MCP server entrypoint
│ ├── coolify-client.ts # Coolify API client
│ └── types.ts # Type definitions
├── dist/ # Compiled JS (generated)
├── .env # Environment variables (create this)
├── setup-mcp.js # Setup script for easy configuration
├── package.json
├── tsconfig.json
└── README.md # This file
API Reference
Selected endpoint mapping used by this server
| MCP Tool | Method | Path |
|---|---|---|
get_version |
GET | /version |
health_check |
GET | /health |
list_teams |
GET | /teams |
get_team |
GET | /teams/{id} |
get_current_team |
GET | /teams/current |
get_current_team_members |
GET | /teams/current/members |
list_applications |
GET | /applications |
create_application |
POST | /applications |
start_application |
POST | /applications/{uuid}/start |
stop_application |
POST | /applications/{uuid}/stop |
restart_application |
POST | /applications/{uuid}/restart |
list_services |
GET | /services |
create_service |
POST | /services |
list_deployments |
GET | /deployments |
get_deployment |
GET | /deployments/{uuid} |
list_private_keys |
GET | /security/keys |
create_private_key |
POST | /security/keys |
get_server_domains |
GET | /servers/{uuid}/domains |
Endpoint paths follow the official Coolify API and do not use a
/api/v1prefix.
Environment variable flags
| Flag | Meaning | Apps | Services |
|---|---|---|---|
is_build_time |
Present at build time | ✅ | — |
is_preview |
Applies to preview deployments only | ✅ | ✅ |
is_literal |
Disable variable substitution | ✅ | ✅ |
Services are prebuilt images (no build phase), so
is_build_timehas no effect for services.
Troubleshooting
Server won't start
- Confirm Node ≥ 18:
node --version - Reinstall deps:
rm -rf node_modules && npm install - Rebuild:
npm run clean && npm run build - Confirm absolute paths in MCP config
Auth errors (401 Unauthorized)
- Verify token is valid and scoped to the correct team
- Recreate token if expired; paste into
.env
Connection issues (ECONNREFUSED/timeouts)
- Check
COOLIFY_BASE_URL(no trailing slash) - Ensure the Coolify instance is reachable and healthy
- Increase
COOLIFY_TIMEOUTfor slow links
Tools not appearing
- Restart your client (Claude Code / Cursor)
- Verify with
claude mcp list - Check server logs in your terminal
- Ensure environment variables are set for the MCP process
Environment variable issues
If you see COOLIFY_BASE_URL and COOLIFY_API_TOKEN environment variables are required errors:
-
Use the automated setup script:
npm run setup -
Or use explicit environment variables: Remove the existing MCP server and re-add with explicit env vars:
claude mcp remove coolify -s local claude mcp add-json coolify '{ "type": "stdio", "command": "node", "args": ["/absolute/path/to/your/dist/index.js"], "env": { "COOLIFY_BASE_URL": "https://your-coolify-instance.com", "COOLIFY_API_TOKEN": "your-api-token-here" } }' -s local -
Verify your .env file exists and has proper values:
cat .env
Why this happens: When MCP servers are launched by Claude Code, they may run from a different working directory, causing the
.envfile to not be found. Explicit environment variables in the MCP configuration solve this issue.
Debug mode
DEBUG=coolify-mcp npm run dev
Security
- Keep API tokens out of VCS; rotate regularly
- Use HTTPS; consider IP allowlists / VPN for private instances
- Add
.envto.gitignore; restrict file permissions - Use least-privilege tokens per environment; audit and revoke regularly
Contributing
We welcome contributions!
-
Fork the repo and clone your fork
git clone https://github.com/<your-username>/Coolify-MCP-Server-for-Claude-Code.git -
Create a branch
git checkout -b feat/your-feature -
Code & test – follow existing style; add tests where useful
-
Commit
git commit -m "feat: add <thing>" -
Push & open PR
git push origin feat/your-feature
Guidelines
- TypeScript best practices; backward compat where feasible
- Clear commit messages and error handling
- Update README when adding features
License
MIT — see LICENSE.
Support
Issues – open on GitHub
Docs
Version: 1.0.0 Maintainer: @forsonny
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。