n8n-workflow-builder（n8n 工作流程构建器）

n8n Workflow Builder

Build production-ready n8n workflows using the available MCP tools.

MCP Tool Decision Tree

Discovery Phase

Configuration Phase

Validation Phase

Deployment Phase (requires API)

Common Workflow Patterns

Pattern 1: Webhook -> Process -> Respond

Webhook -> Code (validate) -> HTTP Request -> Respond

Template: assets/templates/workflows/webhook-processor.json

Pattern 2: Lead Capture -> CRM -> Notify

Webhook -> IF (validate) -> CRM -> Slack -> Rails callback

Template: assets/templates/workflows/lead-capture-crm.json

Pattern 3: AI Agent with Tools

Chat Trigger -> AI Agent <- OpenAI Model
                   ^
                   |- HTTP Request Tool
                   |- Code Tool

Template: assets/templates/workflows/ai-agent-basic.json

Pattern 4: WhatsApp Bot (Israeli Market)

Webhook -> Code (parse) -> Switch (msg type) -> HTTP Request (WhatsApp API)

Template: assets/templates/workflows/whatsapp-bot.json

Pattern 5: Scheduled Reports

Schedule Trigger -> HTTP (fetch data) -> Code (format) -> Email/Slack

Template: assets/templates/workflows/scheduled-report.json

Pattern 6: Rails Integration

Rails App -> Webhook -> Process -> HTTP (callback to Rails)

Template: assets/templates/workflows/rails-webhook-handler.json

MCP Tool Usage Examples

search_nodes - Finding Nodes

// Find database nodes
search_nodes({query: "database"})

// Fuzzy search (typo-tolerant)
search_nodes({query: "slak", mode: "FUZZY"})

// With real examples
search_nodes({query: "webhook", includeExamples: true})

get_node - Node Configuration

// Standard detail (recommended starting point)
get_node({nodeType: "nodes-base.httpRequest", detail: "standard"})

// With real-world examples
get_node({nodeType: "nodes-base.slack", detail: "standard", includeExamples: true})

// Documentation mode
get_node({nodeType: "nodes-base.webhook", mode: "docs"})

// Search specific properties
get_node({nodeType: "nodes-base.httpRequest", mode: "search_properties", propertyQuery: "header"})

// Version comparison
get_node({nodeType: "nodes-base.httpRequest", mode: "compare", fromVersion: "1.0", toVersion: "2.0"})

validate_node - Pre-build Validation

// Full validation with suggestions
validate_node({
  nodeType: "nodes-base.slack",
  config: {resource: "channel", operation: "create"},
  mode: "full",
  profile: "strict"
})

// Quick check
validate_node({
  nodeType: "nodes-base.httpRequest",
  config: {method: "POST", url: "https://api.example.com"},
  mode: "minimal"
})

validate_workflow - Full Workflow Validation

validate_workflow({
  workflow: myWorkflowJson,
  options: {
    validateNodes: true,
    validateConnections: true,
    validateExpressions: true,
    profile: "strict"
  }
})

search_templates - Finding Templates

// By task type
search_templates({searchMode: "by_task", task: "ai_automation"})
search_templates({searchMode: "by_task", task: "webhook_processing"})

// By nodes used
search_templates({searchMode: "by_nodes", nodeTypes: ["n8n-nodes-base.openAi", "n8n-nodes-langchain.agent"]})

// By complexity
search_templates({searchMode: "by_metadata", complexity: "simple", maxSetupMinutes: 15})

// Keyword search
search_templates({query: "slack notification"})

AI Agent Workflows

Connection Types (CRITICAL)

AI connections flow TO the consumer node:

Minimal AI Agent Setup

1. Add Chat Trigger (or Manual Trigger)
2. Add Language Model (OpenAI Chat Model / Anthropic)
3. Add AI Agent
4. Connect: Chat Trigger -> AI Agent (main connection)
5. Connect: Language Model -> AI Agent (ai_languageModel type)

Adding Tools to Agent

// HTTP Request Tool - for API calls
{
  "name": "Call API Tool",
  "type": "@n8n/n8n-nodes-langchain.toolHttpRequest",
  "parameters": {
    "description": "Calls external API to fetch data"  // REQUIRED: 15+ chars
  }
}
// Connect: HTTP Tool -> AI Agent (ai_tool type)

Tool Description Requirements

• Minimum 15 characters
• Must clearly describe what the tool does
• Agent uses this to decide when to use the tool

See references/ai-agents-guide.md for advanced patterns.

Israeli Market Integration

WhatsApp Business API

// Send text message
{
  "method": "POST",
  "url": "https://graph.facebook.com/v18.0/{{PHONE_ID}}/messages",
  "headers": {"Authorization": "Bearer {{ACCESS_TOKEN}}"},
  "body": {
    "messaging_product": "whatsapp",
    "to": "972501234567",
    "type": "text",
    "text": {"body": "\u200F" + hebrewText}  // RTL marker for Hebrew
  }
}

Hebrew RTL Handling

// Add RTL marker for Hebrew text
const rtlText = "\u200F" + hebrewContent;  // Right-to-Left Mark

// Mixed content
const mixed = `\u200F${hebrewPart}\u200E${englishPart}`;  // LTR mark for English

Rate Limits by Tier

See references/israeli-market-integrations.md for WhatsApp templates and payment patterns.

Rails Integration

Triggering n8n from Rails

# app/services/n8n_service.rb
def trigger_webhook(webhook_url, payload)
  HTTParty.post(webhook_url, {
    body: payload.to_json,
    headers: {'Content-Type' => 'application/json', 'X-Webhook-Secret' => ENV['N8N_SECRET']}
  })
end

Receiving n8n Callbacks

# Verify signature
signature = request.headers['X-N8N-Signature']
expected = OpenSSL::HMAC.hexdigest('sha256', ENV['N8N_SECRET'], request.raw_post)
unless ActiveSupport::SecurityUtils.secure_compare(signature.to_s, expected)
  render json: {error: 'Unauthorized'}, status: 401
end

Selling Automations

• Automation model with workflow_json, price, demo_webhook_url
• Purchase model with Stripe integration
• Delivery: JSON download, hosted deploy, or embedded viewer

See references/rails-integration-patterns.md for full implementation.

Validation Checklist

Before Building

• [ ] Run search_nodes to find correct node type
• [ ] Run get_node with detail: "standard" to understand config
• [ ] Check for required credentials

Node Level

• [ ] validate_node({..., mode: "full"}) passes
• [ ] All required fields populated
• [ ] Credentials configured (not hardcoded)
• [ ] Error handling configured

Workflow Level

• [ ] validate_workflow({...}) passes
• [ ] All nodes connected (no orphans)
• [ ] Error Trigger node present
• [ ] Webhook paths unique

AI Workflows

• [ ] Language model connected BEFORE AI Agent
• [ ] All tools have descriptions (15+ chars)
• [ ] Memory node if conversation history needed
• [ ] Streaming: no main outputs from AI Agent

Pre-Deployment

• [ ] Test with sample data
• [ ] Environment variables for sensitive data
• [ ] Rate limiting considered

Workflow JSON Structure

{
  "name": "Workflow Name",
  "nodes": [
    {
      "id": "uuid",
      "name": "Node Name",
      "type": "n8n-nodes-base.webhook",
      "typeVersion": 2,
      "position": [250, 300],
      "parameters": {}
    }
  ],
  "connections": {
    "Node Name": {
      "main": [[{"node": "Next Node", "type": "main", "index": 0}]]
    }
  },
  "settings": {
    "executionOrder": "v1"
  }
}

Scripts

Validate Before Deploy

python scripts/validate_before_deploy.py workflow.json

Generate Workflow Skeleton

python scripts/generate_workflow_skeleton.py --type webhook --name "My Workflow"
python scripts/generate_workflow_skeleton.py --type ai-agent --name "Support Bot"

References

Templates

Workflows

• assets/templates/workflows/webhook-processor.json
• assets/templates/workflows/lead-capture-crm.json
• assets/templates/workflows/whatsapp-bot.json
• assets/templates/workflows/ai-agent-basic.json
• assets/templates/workflows/scheduled-report.json
• assets/templates/workflows/rails-webhook-handler.json

Nodes

• assets/templates/nodes/webhook-trigger.json
• assets/templates/nodes/http-request-auth.json
• assets/templates/nodes/code-transform.json
• assets/templates/nodes/if-conditions.json
• assets/templates/nodes/error-handler.json

Israeli Market

• assets/templates/israel/whatsapp-message-types.json
• assets/templates/israel/hebrew-templates.json
• assets/templates/israel/payment-webhook.json