make-mcp-server
Enables AI assistants to search 200+ Make.com modules, validate and auto-heal blueprints, and deploy automation scenarios directly to Make.com via API.
README
Make.com MCP Server — Build & Deploy Automation Scenarios with AI
⚠️ Disclaimer: This is an unofficial, community-driven project created by a passionate fan of Make.com. It is NOT affiliated with, endorsed by, or officially supported by Make.com.
A Model Context Protocol (MCP) server that gives AI assistants like Claude, GitHub Copilot, and Cursor full access to Make.com module documentation, scenario validation, and one-click deployment. Search 200+ automation modules across 40+ apps, build blueprints with auto-healing, and deploy directly to Make.com — all from your AI chat.
Features
- 🔍 200+ Modules — Full-text search across 200+ Make.com modules (Slack, Gmail, Google Sheets, Notion, OpenAI, and 35+ more apps)
- 📋 Module Details — Retrieve parameters, types, descriptions, and usage docs for any module
- ✅ Blueprint Validation — Check scenarios for missing parameters, unknown modules, structural issues, and router sub-routes before deploying
- 🛡️ Account-Aware Compatibility — Verify module IDs against your live Make account/region before deployment to prevent "Module not found" scenarios
- 🚀 Deploy to Make.com — Push validated blueprints directly to Make.com via API
- 🩹 Auto-Healing — Automatically fixes LLM-generated blueprints: injects missing
metadata, addsdesignercoordinates, strips unsupported properties like routerfilter - 🔀 Router Support — Full support for
builtin:BasicRouterwith multiple routes and recursive validation - 📚 Scenario Templates — Browse reusable scenario templates for common workflows
- 📖 Guided Prompts — MCP prompts for guided scenario building and module exploration
- 📊 Resource Catalog — MCP resources for browsing available apps
- 🧪 43 Tests — Unit + integration test suite with Vitest
- ⚡ Fast Response — Optimized SQLite with FTS5 full-text search
🚀 Quick Start — Self-Hosting Options
Option A: npx (No Installation Needed!) 🚀
The fastest way to get started — no cloning, no building:
Prerequisites: Node.js installed on your system
# Run directly — no installation needed!
npx -y make-mcp-server
The package includes a pre-built database with all 200+ Make.com modules. Just add it to your MCP client config and go.
Claude Desktop config (claude_desktop_config.json):
Basic configuration (documentation tools only):
{
"mcpServers": {
"make-mcp-server": {
"command": "npx",
"args": ["-y", "make-mcp-server"],
"env": {
"LOG_LEVEL": "error"
}
}
}
}
Full configuration (with Make.com deployment):
{
"mcpServers": {
"make-mcp-server": {
"command": "npx",
"args": ["-y", "make-mcp-server"],
"env": {
"LOG_LEVEL": "error",
"MAKE_API_KEY": "your_api_key_here",
"MAKE_TEAM_ID": "your_team_id",
"MAKE_API_URL": "https://eu1.make.com/api/v2"
}
}
}
}
Note: npx will download and cache the latest version automatically. The package includes a pre-built database with all Make.com module information — no setup required.
Option B: Docker (Isolated & Reproducible) 🐳
Prerequisites: Docker installed on your system
# Build the Docker image
docker build -t make-mcp-server .
# Test it works
echo '{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"0.1"}},"id":1}' | docker run -i --rm make-mcp-server
Claude Desktop config:
Basic configuration (documentation tools only):
{
"mcpServers": {
"make-mcp-server": {
"command": "docker",
"args": [
"run", "-i", "--rm", "--init",
"-e", "LOG_LEVEL=error",
"make-mcp-server"
]
}
}
}
Full configuration (with Make.com deployment):
{
"mcpServers": {
"make-mcp-server": {
"command": "docker",
"args": [
"run", "-i", "--rm", "--init",
"-e", "LOG_LEVEL=error",
"-e", "MAKE_API_KEY=your_api_key_here",
"-e", "MAKE_TEAM_ID=your_team_id",
"-e", "MAKE_API_URL=https://eu1.make.com/api/v2",
"make-mcp-server"
]
}
}
}
Important: The
-iflag is required for MCP stdio communication.
Option C: Local Installation (For Development) 🛠️
Prerequisites: Node.js and Git
# 1. Clone and install
git clone https://github.com/danishashko/make-mcp.git
cd make-mcp
npm install
# 2. Build
npm run build
# 3. Populate the module database (already done if using npm package)
npm run scrape:prod
# 4. Test it works
npm start
Claude Desktop config:
{
"mcpServers": {
"make-mcp-server": {
"command": "node",
"args": ["/absolute/path/to/make-mcp/dist/mcp/server.js"],
"env": {
"LOG_LEVEL": "error",
"MAKE_API_KEY": "your_api_key_here",
"MAKE_TEAM_ID": "your_team_id"
}
}
}
}
Note: The Make.com API credentials are optional. Without them, you'll have access to all documentation, search, and validation tools. With them, you'll additionally get scenario deployment capabilities.
Configuration File Locations
| Platform | Path |
|---|---|
| macOS | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Windows | %APPDATA%\Claude\claude_desktop_config.json |
| Linux | ~/.config/Claude/claude_desktop_config.json |
Restart Claude Desktop after updating configuration.
💻 Connect Your IDE
make-mcp-server works with any MCP-compatible client:
- Claude Desktop — See configurations above
- VS Code (GitHub Copilot) — Add to
.vscode/mcp.json - Cursor — Add to MCP settings
- Claude Code — Use
claude mcp addcommand - Windsurf — Add to MCP configuration
Usage
Then ask your AI assistant things like:
"Create a Make scenario that watches a Slack channel for new messages and logs them to a Google Sheet"
"What modules does Make have for sending emails?"
"Validate this scenario blueprint..."
Tip: The AI will automatically call tools_documentation first to understand how to use the server effectively.
Available Tools
| Tool | Description |
|---|---|
tools_documentation |
START HERE — Returns comprehensive documentation for all tools, prompts, and resources |
search_modules |
Full-text search across 200+ Make.com modules |
get_module |
Get detailed module info with parameters and docs |
check_account_compatibility |
Check if modules are available in your current Make account/region (with suggestions) |
validate_scenario |
Validate a scenario blueprint before deployment |
create_scenario |
Deploy a scenario to Make.com via API |
search_templates |
Search reusable scenario templates |
list_apps |
List all apps with module counts |
Auto-Healing
The create_scenario tool automatically fixes common issues in LLM-generated blueprints:
| Issue | Auto-Fix |
|---|---|
Missing metadata section |
Injects full metadata with version, scenario config, and designer |
Missing metadata.designer on modules |
Adds { x: 0, y: 0 } coordinates |
Router filter in route objects |
Strips unsupported filter property (configure filters in Make.com UI) |
Missing version on modules |
Left unset — Make.com auto-resolves the latest installed version |
Catalog mismatch (IM007, module not available) |
Checks live modules in your account/region, auto-remaps close matches, retries deploy once |
Tip: Do NOT hardcode
"version": 1on modules. Some apps (e.g., HTTP) are on v4+ and specifying the wrong version causes "Module not found" errors.
MCP Prompts
| Prompt | Description |
|---|---|
build_scenario |
Guided workflow for creating a Make.com scenario from a natural language description |
explain_module |
Get a detailed explanation of any Make.com module with usage examples |
MCP Resources
| Resource URI | Description |
|---|---|
make://apps |
List of all available apps with module counts |
CLI Usage
make-mcp-server # Start the MCP server (stdio transport)
make-mcp-server --scrape # Populate/refresh the module database
make-mcp-server --version # Print version
make-mcp-server --help # Show help
Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
MAKE_API_KEY |
For deployment | — | Make.com API key |
MAKE_API_URL |
No | https://eu1.make.com/api/v2 |
Make.com API base URL |
MAKE_TEAM_ID |
For deployment | — | Default team ID for scenario deployment |
DATABASE_PATH |
No | <package>/data/make-modules.db |
SQLite database file path |
LOG_LEVEL |
No | info |
Logging level: debug, info, warn, error, silent |
MAKE_MODULE_CACHE_TTL_MS |
No | 300000 |
Cache TTL for live module catalog checks (milliseconds) |
Development
npm run build # Compile TypeScript + copy schema + add shebang
npm run build:tsc # TypeScript only (no packaging)
npm run start:dev # Start with tsx (no build needed)
npm run dev # Start with file watching
npm run scrape # Populate DB with tsx (dev)
npm run scrape:prod # Populate DB from compiled JS
npm run smoke:compat # One-command account compatibility + validation smoke check
npm test # Run all 43 tests
npm run test:watch # Run tests in watch mode
Publishing to npm
npm run prepublishOnly # Build + populate DB + verify (runs automatically on npm publish)
npm publish # Publish to npm registry
Testing
The test suite includes 43 tests across 3 files:
- Database tests (14 tests) — Insert, search, template operations, FTS5 queries
- Logger tests (7 tests) — Stderr-only output, log levels, data serialization
- Server integration tests (22 tests) — Full MCP protocol compliance via SDK client
npm test
Fast Verification (2-3 minutes)
Use this when you just changed validation/deploy logic and want confidence quickly:
# 1) Fast regression check
npm test
# 2) One-command smoke check (compatibility + validation)
npm run smoke:compat
# 3) (Optional) Manual MCP exploration
npm run start:dev
Then in your MCP client (Claude/Cursor/Copilot), run this sequence:
check_account_compatibilitywith a known module, e.g.gateway:CustomWebHookvalidate_scenariowith your blueprintcreate_scenarioonly if compatibility is good
If step 1 reports incompatible modules, use the suggested replacement IDs before deploying.
Architecture
src/
├── mcp/
│ └── server.ts # MCP server with tools, prompts, resources
├── database/
│ ├── schema.sql # SQLite + FTS5 schema
│ └── db.ts # Database access layer (npx-safe path resolution)
├── scrapers/
│ └── scrape-modules.ts # Module data population (224 modules)
└── utils/
└── logger.ts # Structured stderr-only logger
bin/
├── make-mcp.js # CLI entry point (npx, --help, --version, --scrape)
└── postinstall.js # Post-install verification
scripts/
├── build.js # Build: tsc + copy schema + add shebang
├── prepublish.js # Publish prep: build + populate DB + verify
└── smoke-compat.js # Fast local smoke test for module compatibility + validation
data/
└── make-modules.db # Pre-built SQLite database (bundled in npm package)
tests/
├── database.test.ts # Database unit tests (14)
├── logger.test.ts # Logger unit tests (7)
└── server.test.ts # MCP integration tests (22)
Dockerfile # Multi-stage Docker image
Tech Stack
- TypeScript + Node.js (ESM)
- @modelcontextprotocol/sdk v1.26.0 — MCP protocol implementation
- better-sqlite3 — Synchronous SQLite with FTS5 full-text search
- zod — Schema validation for tool parameters
- axios — HTTP client for Make.com API
- vitest — Test framework
Supported Apps (40+)
Google Sheets, Slack, OpenAI, Google Drive, Notion, Telegram Bot, HubSpot CRM, Gmail, Airtable, Tools, Flow Control, Google Calendar, Jira, Trello, Shopify, Google Docs, Microsoft Teams, Microsoft Outlook, Discord, Asana, monday.com, Salesforce, Stripe, GitHub, HTTP, Mailchimp, WordPress, Dropbox, Data Store, JSON, Twilio, Google Gemini AI, WhatsApp Business, Text Parser, Webhooks, Anthropic Claude, CSV, RSS, Email, Schedule
Author
Built by Daniel Shashko
License
MIT License — see LICENSE for details.
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。