[![npm version](https://img.shields.io/npm/v/flowengine-n8n-workflow-builder)](https://www.npmjs.com/package/flowengine-n8n-workflow-builder)
[![Smithery](https://img.shields.io/badge/Smithery-Install-blue)](https://smithery.ai/server/@Ami3466/mcp-flowengine-n8n-workflow-builder)
# n8n Workflow Builder MCP Server

**Build n8n workflows from text using AI** - Built by [FlowEngine](https://flowengine.cloud)

Turn natural language into **production-ready, validated** n8n workflows. This isn't just another n8n context provider, it's a complete workflow generation engine with built-in validation, auto-fixing, and architectural intelligence.

**this MCP is built to provide validated n8n workflows — not just context.**

- **13-Point Validation Engine** - Catches errors before you import
- **Auto-Fix Malformed Workflows** — Automatically repairs common issues
- **Architecture Recommendations** — Suggests optimal workflow patterns
- **Security Scanning** — Detects credential leaks and vulnerabilities
- **Performance Analysis** — Identifies bottlenecks and optimization opportunities
- **Real Parameter Schemas** — Loaded directly from n8n packages for accuracy
- **600+ Registered Node Types** — Only real nodes, no hallucinations

**Result:** Workflows that actually import and run on first try.

## Quick Start

**[Local Install](#setup-by-platform)** | **[Remote Server (no install)](#option-b-remote-server-more-stable---unlimited-)**

## Video Demo

[![Watch the demo](https://img.youtube.com/vi/XrxHom6w6WM/maxresdefault.jpg)](https://youtu.be/XrxHom6w6WM)

## Feature Details

### 13-Point Validation Engine

Every workflow passes through 13 validation checks before output:

1. **Node Type Validation** — Verifies every node type exists in n8n
2. **Connection Integrity** — Ensures all connections reference existing nodes
3. **Parameter Type Checking** — Validates parameter types match node schemas
4. **Required Fields** — Checks all required parameters are present
5. **Credential References** — Validates credential configurations
6. **Expression Syntax** — Checks n8n expression syntax (`={{ }}`)
7. **Position Validation** — Ensures nodes have valid canvas positions
8. **Duplicate Detection** — Catches duplicate node names
9. **Orphan Node Detection** — Finds disconnected nodes
10. **Trigger Validation** — Ensures workflows have proper entry points
11. **Loop Detection** — Identifies potential infinite loops
12. **Output Mapping** — Validates data flow between nodes
13. **Version Compatibility** — Checks node version compatibility

### Auto-Fix Malformed Workflows

When validation finds issues, the engine automatically repairs them:

- **Missing positions** → Auto-calculates layout on canvas
- **Invalid node names** → Generates unique, valid names
- **Missing connections array** → Initializes proper structure
- **Incorrect typeVersion** → Updates to current supported version
- **Malformed parameters** → Applies sensible defaults
- **Broken JSON structure** → Attempts recovery and repair

### Architecture Recommendations

Based on your task description, the engine analyzes keywords and intent to suggest the optimal pattern. Here's when each is recommended:

#### Regular Workflows (Deterministic)

**Linear Pipeline**
- **When:** Simple, predictable data transformations
- **Example:** "Get data from API → Transform → Save to database"
- **Best for:** ETL, data sync, scheduled reports
- **Why not AI:** No decision-making needed, faster execution, lower cost

**Conditional Branching**
- **When:** Known decision points with clear rules
- **Example:** "If order > $100, send to manager; else auto-approve"
- **Best for:** Approval flows, routing, rule-based automation
- **Why not AI:** Rules are explicit, no reasoning required

**Parallel Processing**
- **When:** Independent operations that can run simultaneously
- **Example:** "Send email AND update CRM AND log to Slack"
- **Best for:** Notifications, multi-system updates, batch processing
- **Why not AI:** No dependencies between branches

**Event-Driven**
- **When:** Reacting to external triggers
- **Example:** "When webhook received → process → respond"
- **Best for:** API endpoints, real-time integrations, chatbots
- **Why not AI:** Response is formulaic, not conversational

#### AI Agent Workflows (Autonomous)

**When to use AI Agents instead of regular workflows:**

| Use AI Agent When... | Use Regular Workflow When... |
|---------------------|------------------------------|
| Task requires reasoning | Steps are predictable |
| Input is unstructured (natural language) | Input is structured (JSON, forms) |
| Multiple tools might be needed dynamically | Tool sequence is known |
| Conversation/context matters | Stateless processing |
| Decision logic is complex or fuzzy | Rules are explicit |

**Single AI Agent**
- **When:** One autonomous entity with access to tools
- **Example:** "Customer support bot that can search docs, create tickets, and escalate"
- **Architecture:** Chat Trigger → AI Agent (with tools) → Response
- **Tools:** Calculator, Code, HTTP Request, custom tools
- **Memory:** Optional conversation memory for context

**AI Agent with Memory**
- **When:** Conversation context matters across messages
- **Example:** "Personal assistant that remembers user preferences"
- **Architecture:** Chat Trigger → Memory Load → AI Agent → Memory Save → Response
- **Memory types:** Buffer (recent), Window (last N), Summary (compressed)

**Multi-Agent Workflows**

For complex tasks, multiple specialized agents can collaborate:

**Sequential Multi-Agent**
- **When:** Task has distinct phases requiring different expertise
- **Example:** "Research Agent → Analysis Agent → Writing Agent"
- **Flow:** Agent 1 output becomes Agent 2 input
- **Best for:** Content pipelines, multi-step reasoning

**Supervisor Pattern**
- **When:** Need coordination between specialist agents
- **Example:** "Supervisor routes to: Sales Agent, Support Agent, or Technical Agent"
- **Flow:** Supervisor Agent decides which specialist handles the task
- **Best for:** Customer service, complex routing

**Parallel Agents**
- **When:** Multiple perspectives needed simultaneously
- **Example:** "Analyst Agent AND Risk Agent AND Compliance Agent all review"
- **Flow:** Split → Multiple Agents → Merge results
- **Best for:** Review processes, multi-criteria evaluation

**Hierarchical Agents**
- **When:** Complex orchestration with sub-tasks
- **Example:** "Manager Agent delegates to Team Agents who use Tool Agents"
- **Flow:** Top-level agent breaks down task, delegates, aggregates
- **Best for:** Large-scale automation, enterprise workflows

#### How the Engine Decides

The recommendation engine analyzes your description for:

1. **AI keywords:** "chat", "conversation", "understand", "decide", "reason" → suggests AI Agent
2. **Automation keywords:** "sync", "transform", "schedule", "trigger" → suggests Regular Workflow
3. **Complexity signals:** Multiple conditions, dynamic routing → suggests Conditional or Agent
4. **Tool mentions:** "search", "calculate", "call API" → adds appropriate tools to Agent
5. **Memory signals:** "remember", "context", "history" → adds memory to Agent

### Security Scanning

Detects vulnerabilities before deployment:

- **Hardcoded credentials** — API keys, passwords, tokens in plain text
- **Insecure protocols** — HTTP instead of HTTPS
- **Missing authentication** — API calls without auth headers
- **Sensitive data exposure** — PII in logs or outputs
- **Code injection risks** — Unsafe code node patterns

### Performance Analysis

Identifies bottlenecks and optimization opportunities:

- **Execution time estimates** — Per-node and total workflow
- **Parallel path detection** — Opportunities for concurrent execution
- **API call optimization** — Batch vs. individual requests
- **Memory usage patterns** — Large data handling concerns
- **Rate limiting risks** — High-frequency API calls

### Real Parameter Schemas

Parameters are loaded directly from n8n packages, not guessed:

- **Accurate defaults** — Real default values from node definitions
- **Correct types** — String, number, boolean, options validated
- **Required vs optional** — Knows which fields are mandatory
- **Nested structures** — Complex parameter objects handled correctly

### 600+ Registered Node Types

**No hallucinations.** The AI can only use nodes that actually exist:

- Every node type is loaded from the official `n8n-nodes-base` package
- LangChain nodes from `@n8n/n8n-nodes-langchain` included
- Node registry is generated directly from n8n source code
- If a node doesn't exist in the registry, it won't be used
- Prevents the common AI problem of inventing fake node types


---

## Quick Start

### Installation Methods

Choose the method that works best for you:

#### 1. Smithery (One-Click Install)

Easy local installation for Claude Desktop. [View on Smithery](https://smithery.ai/server/@Ami3466/mcp-flowengine-n8n-workflow-builder)

```bash
npx @smithery/cli install @Ami3466/mcp-flowengine-n8n-workflow-builder --client claude
```

**Note:** Smithery has a 50 uses/day limit. For unlimited access, use the remote server below.

#### 2. Remote Server (More Stable - Unlimited) ⭐

**Connect to our hosted server - no installation required!**

Add this to your Claude Desktop config:

**Mac:** `~/Library/Application Support/Claude/claude_desktop_config.json`
**Windows:** `%APPDATA%/Claude/claude_desktop_config.json`

```json
{
  "mcpServers": {
    "flowengine-n8n": {
      "url": "https://mcp-flowengine-n8n-workflow-builder.onrender.com/mcp",
      "transport": "http"
    }
  }
}
```

**Benefits:**
- ✅ **More stable** - Dedicated hosting infrastructure
- ✅ **Unlimited usage** - No daily limits
- ✅ **Always up-to-date** - Automatically updated
- ✅ **Zero installation** - Just add config and restart Claude
- ✅ **Free** - Community hosted

#### 3. npm (Manual Install)

For all other MCP clients:

```bash
npm install -g flowengine-n8n-workflow-builder
```

---

## Setup by Platform

### Claude Desktop

#### Option A: Smithery (Local Install)

[View on Smithery](https://smithery.ai/server/@Ami3466/mcp-flowengine-n8n-workflow-builder)

```bash
npx @smithery/cli install @Ami3466/mcp-flowengine-n8n-workflow-builder --client claude
```

Restart Claude Desktop after installation.

**Note:** Limited to 50 uses/day. For unlimited access, use the remote server below.

#### Option B: Remote Server (More Stable - Unlimited) ⭐

**No installation required!** Just add config and restart.

1. **Edit Claude Desktop config:**
   - **Mac:** `~/Library/Application Support/Claude/claude_desktop_config.json`
   - **Windows:** `%APPDATA%/Claude/claude_desktop_config.json`

2. **Add this configuration:**
   ```json
   {
     "mcpServers": {
       "flowengine-n8n": {
         "url": "https://mcp-flowengine-n8n-workflow-builder.onrender.com/mcp",
         "transport": "http"
       }
     }
   }
   ```

3. **Restart Claude Desktop** (fully quit and reopen)

4. **Verify connection:**
   - Look for the 🔌 MCP icon in Claude Desktop
   - You should see "flowengine-n8n" listed with green status

**Benefits:** More stable, unlimited usage, no local installation, always updated.

#### Option C: Manual Local Install

1. **Install the package:**
   ```bash
   npm install -g flowengine-n8n-workflow-builder
   ```

2. **Edit Claude Desktop config:**
   - **Mac:** `~/Library/Application Support/Claude/claude_desktop_config.json`
   - **Windows:** `%APPDATA%/Claude/claude_desktop_config.json`

3. **Add this configuration:**
   ```json
   {
     "mcpServers": {
       "flowengine-n8n": {
         "command": "flowengine-n8n"
       }
     }
   }
   ```

4. **Restart Claude Desktop** (fully quit and reopen)

### Claude Code (VS Code Extension)

1. **Install the package globally:**
   ```bash
   npm install -g flowengine-n8n-workflow-builder
   ```

2. **Open VS Code Settings** (Cmd/Ctrl + ,)

3. **Search for "MCP"**

4. **Add MCP Server:**
   - Click "Edit in settings.json"
   - Add to `claude.mcpServers`:
   ```json
   {
     "claude.mcpServers": {
       "flowengine-n8n": {
         "command": "flowengine-n8n"
       }
     }
   }
   ```

5. **Reload VS Code** (Cmd/Ctrl + Shift + P → "Developer: Reload Window")

6. **Start using:**
   - Open Claude Code panel
   - Ask Claude to build n8n workflows
   - The MCP server will be automatically available

### Cursor

1. **Install the package:**
   ```bash
   npm install -g flowengine-n8n-workflow-builder
   ```

2. **Open Cursor Settings:**
   - **Mac:** Cursor → Settings → Features
   - **Windows/Linux:** File → Preferences → Features

3. **Navigate to MCP Settings:**
   - Scroll to "Model Context Protocol"
   - Click "Edit Config"

4. **Add configuration:**
   ```json
   {
     "mcpServers": {
       "flowengine-n8n": {
         "command": "flowengine-n8n"
       }
     }
   }
   ```

5. **Restart Cursor**

6. **Verify:**
   - Open Cursor's AI chat
   - The MCP server should be available
   - Try: "Create an n8n workflow for me"

### Cline (VS Code)

1. **Install the package:**
   ```bash
   npm install -g flowengine-n8n-workflow-builder
   ```

2. **Open Cline Settings in VS Code:**
   - Open Command Palette (Cmd/Ctrl + Shift + P)
   - Type "Cline: Open Settings"

3. **Add MCP Server:**
   - In Cline settings, find "MCP Servers"
   - Add new server:
   ```json
   {
     "flowengine-n8n": {
       "command": "flowengine-n8n"
     }
   }
   ```

4. **Reload VS Code**

### Continue.dev

1. **Install the package:**
   ```bash
   npm install -g flowengine-n8n-workflow-builder
   ```

2. **Open Continue config:**
   - Open Command Palette (Cmd/Ctrl + Shift + P)
   - Type "Continue: Open config.json"

3. **Add MCP server to config:**
   ```json
   {
     "mcpServers": {
       "flowengine-n8n": {
         "command": "flowengine-n8n"
       }
     }
   }
   ```

4. **Reload Continue extension**

---

## What This Does

### 🤖 Turns Text into Workflows

Describe what you want in plain language:
- "Monitor my email and notify me on Slack"
- "Build an AI chatbot with memory and tools"
- "Sync data between Google Sheets and my database"
- "Create a customer support automation workflow"

Your AI assistant generates complete, working n8n workflows.

### 🧠 Expert n8n Knowledge

Your AI gets access to:
- **600+ Node Types** - All n8n-nodes-base and LangChain nodes
- **Real Parameter Schemas** - Loaded directly from n8n packages
- **Best Practices** - Workflow design patterns and optimizations
- **Intelligent Validation** - Automatic error detection and fixing

### ✨ Powerful Features

**Workflow Generation:**
- Build workflows from natural language descriptions
- Add, edit, and delete nodes programmatically
- Connect nodes and manage workflow structure
- Get detailed workflow analysis

**Intelligence & Suggestions:**
- Architecture recommendations (linear, conditional, AI agent, etc.)
- Node suggestions for specific tasks
- Workflow analysis and improvements
- Natural language explanations

**Quality & Security:**
- Comprehensive workflow validation
- Security vulnerability scanning
- Performance analysis and bottleneck detection
- Dry-run workflow testing

**Templates & Search:**
- Pre-built workflow templates
- Search 600+ nodes by keyword
- Browse nodes by category
- Get detailed node documentation

---

## Usage Examples

Once installed, ask your AI assistant to help with workflows:

### Create a New Workflow

```
"Create an n8n workflow that monitors Gmail for emails with 'urgent' in the subject and sends a Slack notification to #alerts"
```

### Analyze Existing Workflow

```
"Analyze this workflow and suggest improvements"
[paste your workflow JSON]
```

### Get Node Recommendations

```
"What nodes should I use to build a customer onboarding automation?"
```

### Validate and Fix

```
"Validate this workflow and fix any errors"
[paste workflow JSON]
```

### Security Scan

```
"Scan this workflow for security issues"
[paste workflow JSON]
```

---

## Available Tools

Your AI assistant gets access to 23 powerful tools:

**Workflow Building:**
- `build_workflow` - Generate workflows from descriptions
- `add_node` - Add nodes to existing workflows
- `edit_node` - Modify node parameters
- `delete_node` - Remove nodes
- `add_connection` - Connect nodes
- `get_workflow_details` - Analyze workflow structure

**Validation & Quality:**
- `validate_workflow` - Comprehensive validation with auto-fix
- `test_workflow` - Dry-run simulation
- `scan_security` - Security vulnerability detection
- `analyze_performance` - Performance and bottleneck analysis

**Intelligence:**
- `suggest_architecture` - Recommend workflow patterns
- `suggest_nodes` - Node recommendations for tasks
- `analyze_workflow` - Deep workflow insights
- `suggest_improvements` - Optimization suggestions
- `explain_workflow` - Natural language explanations

**Node Library:**
- `search_nodes` - Search 600+ nodes
- `list_nodes_by_category` - Browse by category
- `get_node_details` - Detailed node documentation

**Templates:**
- `list_templates` - Browse workflow templates
- `get_template` - Get specific templates
- `search_templates` - Search templates

**Utilities:**
- `extract_workflow_json` - Extract JSON from text
- `fix_json` - Repair malformed JSON

---

## Deploy Your Workflows

### Option 1: FlowEngine.cloud (Recommended)

**Build for free, deploy instantly:**

1. Generate workflow using this MCP server
2. Visit [flowengine.cloud](https://flowengine.cloud)
3. Import your workflow JSON
4. Test and deploy - no infrastructure needed

**Why FlowEngine.cloud?**
- ✅ No server setup or management
- ✅ Built-in monitoring and logs
- ✅ Automatic scaling
- ✅ Visual workflow editor
- ✅ Free tier available

### Option 2: Self-Hosted n8n

1. Generate workflow using this MCP server
2. Open your n8n instance
3. Import JSON (`...` → `Import from File`)
4. Configure credentials and activate

---

## Troubleshooting

### MCP Server Not Showing Up?

1. **Verify installation:**
   ```bash
   which flowengine-n8n
   # Should show: /usr/local/bin/flowengine-n8n (or similar)
   ```

2. **Test the server manually:**
   ```bash
   flowengine-n8n
   # Should start the MCP server
   ```

3. **Check Claude Desktop logs:**
   - **Mac:** `~/Library/Logs/Claude/mcp*.log`
   - **Windows:** `%APPDATA%/Claude/logs/mcp*.log`

4. **Restart your AI client completely** (fully quit and reopen)

### Package Not Found?

```bash
# Update npm
npm install -g npm@latest

# Reinstall the package
npm uninstall -g flowengine-n8n-workflow-builder
npm install -g flowengine-n8n-workflow-builder
```

### Permission Issues?

**Mac/Linux:**
```bash
sudo npm install -g flowengine-n8n-workflow-builder
```

**Windows:**
Run PowerShell/CMD as Administrator

### Smithery Installation Issues?

```bash
# Clear npm cache
npm cache clean --force

# Try manual installation instead
npm install -g flowengine-n8n-workflow-builder
```

---

## How It Works

This MCP server connects your AI assistant to expert n8n knowledge:

```
You describe what you want
         ↓
Your AI Tool (Claude/Cursor/etc.)
         ↓
   MCP Protocol
         ↓
FlowEngine n8n Builder
         ↓
Expert n8n Knowledge + Validation
         ↓
Complete, Working Workflow
```

**What Your AI Gets:**
- Deep n8n expertise
- 600+ node definitions with real schemas
- Workflow patterns and best practices
- Validation and auto-fixing capabilities
- Security and performance analysis

**What You Get:**
- Production-ready workflows
- Properly validated JSON
- Working configurations
- Best practice implementations

---

## Features

✅ **Build workflows from text** - Natural language to working n8n workflows
✅ **600+ Node Types** - Full n8n-nodes-base and LangChain support
✅ **Real Parameter Schemas** - Loaded directly from n8n packages
✅ **Intelligent validation** - Automatic error detection and fixing
✅ **Security scanning** - Detect vulnerabilities and sensitive data
✅ **Performance analysis** - Find bottlenecks and optimize
✅ **Works with any LLM** - Universal MCP protocol support
✅ **No API keys needed** - Works completely offline
✅ **Built by FlowEngine** - Production-tested technology
✅ **Deploy anywhere** - FlowEngine.cloud or self-hosted

---

## About FlowEngine

**FlowEngine** is a platform for building and deploying n8n workflows with AI:

- **🎨 Visual Builder** - Drag-and-drop editor at [flowengine.cloud](https://flowengine.cloud)
- **🤖 AI-Powered** - Generate workflows with natural language
- **☁️ Managed Hosting** - Deploy instantly, no DevOps needed
- **📈 Production Ready** - Monitoring, logs, and scaling included
- **🆓 Free Tier** - Try and build workflows for free

This MCP server brings FlowEngine's workflow generation technology to your local development environment.

---

## Use Cases

**For Developers:**
- Generate boilerplate workflows quickly
- Prototype automation ideas fast
- Learn n8n patterns and best practices
- Build complex workflows with AI assistance

**For Teams:**
- Accelerate workflow development
- Standardize workflow patterns
- Reduce learning curve for n8n
- Scale automation initiatives

**For Businesses:**
- Automate repetitive tasks
- Connect systems and tools
- Build custom integrations
- Deploy AI-powered workflows

---

## Support & Resources

- **FlowEngine Platform:** [flowengine.cloud](https://flowengine.cloud)
- **Documentation:** [docs.flowengine.cloud](https://docs.flowengine.cloud)
- **npm Package:** [View on npm](https://www.npmjs.com/package/flowengine-n8n-workflow-builder)
- **Smithery Registry:** [Install via Smithery](https://smithery.ai/server/@Ami3466/mcp-flowengine-n8n-workflow-builder)
- **GitHub Repository:** [View Source](https://github.com/Ami3466/flowengine-mcp-n8n-workflow-builder)
- **Report Issues:** Contact FlowEngine support

---

## License

**MIT License with Commons Clause**

✅ **You CAN:**
- Use for personal projects (free forever)
- View and study the source code
- Modify for your own personal use
- Contribute improvements back

❌ **You CANNOT:**
- Sell this software
- Offer as a commercial/paid service
- Build competing products using this code
- Use in any commercial product or service

For commercial licensing: [flowengine.cloud](https://flowengine.cloud)

Full license: [LICENSE](LICENSE)

---

**Built by [FlowEngine](https://flowengine.cloud)** - Enterprise-grade n8n workflow automation platform

[Website](https://flowengine.cloud) • [Documentation](https://docs.flowengine.cloud) • [npm](https://www.npmjs.com/package/flowengine-n8n-workflow-builder) • [Smithery](https://smithery.ai/server/@Ami3466/mcp-flowengine-n8n-workflow-builder)