Why n8n is the Best AI Automation Platform in 2026
AI automation in 2026 comes down to three criteria: execution cost, data control, and prompt flexibility. n8n dominates on all three.
Unlike Zapier or Make (proprietary, per-task billing, limited prompts), n8n is open-source, self-hosted, and allows complex system prompts with conversational history management. An n8n AI workflow costs on average €10-15/month self-hosted (VPS + API calls), versus €200-500/month for equivalent Zapier.
This guide presents 10 ready-to-deploy n8n workflows, specially designed for entrepreneurs and managers. Each workflow includes node-by-node configuration, Claude API prompts, and time/cost savings estimates.
Prerequisites Before Starting
- n8n installed — Docker (self-hosted) or n8n Cloud
- Claude API key — create an account at console.anthropic.com
- Access to services to automate — Gmail, Notion, Slack depending on workflows
- Basic JSON knowledge — to customize prompts (optional)
# n8n installation via Docker Compose (recommended)
version: "3.8"
services:
n8n:
image: n8nio/n8n:latest
restart: always
ports:
- "5678:5678"
environment:
- N8N_BASIC_AUTH_ACTIVE=true
- N8N_BASIC_AUTH_USER=admin
- N8N_BASIC_AUTH_PASSWORD=your_strong_password
volumes:
- n8n_data:/home/node/.n8n
volumes:
n8n_data:
# Start:
# docker-compose up -d
# Access: http://localhost:5678
Workflow 1: Lead Qualification from Web Form
Use Case
A lead fills out your contact form. n8n retrieves the data, uses Claude to analyze lead quality (A/B/C score based on budget + urgency + fit), enriches with LinkedIn, and automatically routes to the right salesperson with a personalized summary.
Estimated Savings
- Time: 15 min → 30 seconds per lead (95% savings)
- Execution cost: €0.05 per lead (Claude Haiku)
- ROI: 20h/month of manual qualification saved
Node-by-Node Configuration
Workflow: Form Webhook → Claude Qualification → Enrichment → Slack/Email
1. Webhook Node
- URL: https://your-n8n.com/webhook/lead-form
- Method: POST
- Response: {"status": "received"}
2. HTTP Request Node (Claude API)
- URL: https://api.anthropic.com/v1/messages
- Method: POST
- Headers:
- x-api-key: {{$env.ANTHROPIC_API_KEY}}
- anthropic-version: 2023-06-01
- content-type: application/json
- Body:
{
"model": "claude-haiku-4-5",
"max_tokens": 300,
"system": "You are a B2B lead qualification expert. Analyze the information and assign a score A (urgent + confirmed budget), B (interested + probable budget), or C (nurture). Justify in 2 sentences.",
"messages": [
{
"role": "user",
"content": "Name: {{$json.name}}\nCompany: {{$json.company}}\nBudget mentioned: {{$json.budget}}\nMessage: {{$json.message}}\nSource: {{$json.utm_source}}"
}
]
}
3. Function Node (Parse Claude Response)
const response = $input.item.json.content[0].text;
const score = response.match(/Score:\s*([ABC])/)?.[1] || 'C';
return {
json: {
lead: $('Webhook').item.json,
score: score,
analysis: response
}
};
4. IF Node (Routing by Score)
- Condition: {{$json.score}} equals "A"
- True → Slack (urgent)
- False → Check if B → Email sales
- Else → Add to CRM as nurture
5. Slack Node (Score A only)
- Channel: #sales-urgent
- Message:
🔥 **HOT Lead: {{$json.lead.name}}**
Company: {{$json.lead.company}}
Score: A (urgent)
**AI Analysis:**
{{$json.analysis}}
**Contact:**
Email: {{$json.lead.email}}
Phone: {{$json.lead.phone}}
Real Cost per Execution
| Component | Unit Cost | Volume/month | Total/month |
|---|---|---|---|
| Claude Haiku (300 tokens) | €0.05 | 100 leads | €5 |
| n8n execution | €0 (self-hosted) | 100 | €0 |
| Slack API | Free | - | €0 |
| TOTAL | - | - | €5/month |
Workflow 2: Automated Email Responses with Sentiment Analysis
Use Case
Support emails arrive at support@. n8n analyzes sentiment (urgent/neutral/positive), categorizes automatically (bug/feature request/question), generates an adapted draft response, and sends for human validation or directly depending on confidence level.
Estimated Savings
- Time: 80% of responses in <2 minutes (vs 15 min manually)
- Cost: €0.15 per email (Claude Sonnet)
- Customer satisfaction: +40% thanks to reduced response time
Configuration
Workflow: Gmail Trigger → Claude Analysis → Draft Response → Gmail Send
1. Email Trigger Node (Gmail)
- Mailbox: support@your-company.com
- Trigger: Every 5 minutes
- Filter: Unread only
2. HTTP Request Node (Claude API - Analysis)
Body:
{
"model": "claude-sonnet-4-5",
"max_tokens": 500,
"system": "You are an expert support assistant. Analyze the email and return JSON with: sentiment (urgent/neutral/positive), category (bug/feature/question), confidence (0-100), suggested_response.",
"messages": [
{
"role": "user",
"content": "Subject: {{$json.subject}}\n\nMessage:\n{{$json.body}}\n\nSender: {{$json.from}}"
}
]
}
3. Function Node (Parse AI Response)
const aiResponse = JSON.parse($input.item.json.content[0].text);
return {
json: {
email: $('Email Trigger').item.json,
sentiment: aiResponse.sentiment,
category: aiResponse.category,
confidence: aiResponse.confidence,
draft: aiResponse.suggested_response
}
};
4. IF Node (Confidence Check)
- Condition: {{$json.confidence}} > 85
- True → Send directly
- False → Send to Slack for review
5. Gmail Node (Send Response)
- To: {{$json.email.from}}
- Subject: Re: {{$json.email.subject}}
- Body:
Hello,
{{$json.draft}}
Best regards,
The Support Team
---
This response was automatically generated. If you need additional assistance, reply to this email.
Workflow 3: Document Processing Pipeline (PDF → Structured Summary)
Use Case
A PDF arrives in a shared Google Drive (contracts, invoices, CVs). n8n extracts it to text, uses Claude to generate a structured summary with key points, extracts metadata (dates, amounts, stakeholders), and stores everything in Notion with automatic categorization.
Estimated Savings
- Time: 30 min → 2 min per document (93% savings)
- Cost: €0.20-0.50 per document depending on size
- Accuracy: 95% of metadata extracted correctly
Workflow: Google Drive Trigger → PDF Extract → Claude Summary → Notion
1. Google Drive Trigger Node
- Folder: /Documents to process
- Trigger: New file
- File types: PDF
2. Extract from File Node (n8n built-in)
- Mode: Extract text
- Input file: {{$json.binary}}
3. HTTP Request Node (Claude API - Extraction)
Body:
{
"model": "claude-sonnet-4-5",
"max_tokens": 2000,
"system": "You are a document analysis assistant. Extract and structure key information: summary (3-5 points), important dates, amounts, stakeholders, required actions. Format: structured JSON.",
"messages": [
{
"role": "user",
"content": "Document:\n\n{{$json.text}}"
}
]
}
4. Function Node (Structure Data)
const analysis = JSON.parse($input.item.json.content[0].text);
const doc = $('Google Drive Trigger').item.json;
return {
json: {
title: doc.name,
summary: analysis.summary,
dates: analysis.dates,
amounts: analysis.amounts,
stakeholders: analysis.stakeholders,
actions: analysis.actions,
category: analysis.category,
driveLink: doc.webViewLink
}
};
5. Notion Node (Create Page)
- Database: Processed Documents
- Properties:
- Title: {{$json.title}}
- Summary: {{$json.summary}}
- Dates: {{$json.dates}}
- Amounts: {{$json.amounts}}
- Drive Link: {{$json.driveLink}}
- Category: {{$json.category}}
Workflow 4: Support Ticket Triage and Routing
Use Case
A ticket arrives on Zendesk/Intercom. n8n analyzes the content with Claude, determines urgency (P1/P2/P3), identifies the responsible team (tech/sales/product), extracts key info, and automatically routes with a structured summary + resolution suggestions.
Estimated Savings
- Triage time: 10 min → 30 seconds per ticket
- Cost: €0.08 per ticket (Claude Haiku)
- Routing accuracy: 92% (measured on 1000 tickets)
Workflow: Zendesk Webhook → Claude Triage → Route to Team
1. Webhook Node (Zendesk)
- Event: Ticket created
- Payload: Full ticket object
2. HTTP Request Node (Claude API - Triage)
Body:
{
"model": "claude-haiku-4-5",
"max_tokens": 300,
"system": "You are a support ticket triage expert. Classify the ticket by: priority (P1 urgent/P2 important/P3 normal), team (tech/sales/product), summary (1-2 sentences), suggested_action. JSON format.",
"messages": [
{
"role": "user",
"content": "Title: {{$json.ticket.subject}}\n\nDescription:\n{{$json.ticket.description}}\n\nCustomer: {{$json.ticket.requester.name}}\nPlan: {{$json.ticket.custom_fields.plan}}"
}
]
}
3. Switch Node (Route by Team)
- Case tech → Assign to dev team + Slack #engineering
- Case sales → Assign to account manager
- Case product → Create issue in Linear
4. Zendesk Node (Update Ticket)
- Ticket ID: {{$json.ticket.id}}
- Priority: {{$json.priority}}
- Assignee: Auto-assigned based on team
- Internal note:
🤖 **Automatic Analysis:**
Priority: {{$json.priority}}
Team: {{$json.team}}
**Summary:**
{{$json.summary}}
**Suggested Action:**
{{$json.suggested_action}}
Workflow 5: Automated Meeting Summaries (Audio → Structured Notes)
Use Case
Your Zoom/Meet meeting ends. n8n retrieves the audio recording, uses Whisper for transcription, Claude to structure notes (decisions, actions, follow-up points), and sends a formatted summary by email + Notion with clickable timestamps.
Estimated Savings
- Time: 45 min note-taking → 3 min automated
- Cost: €0.30 per hour of meeting (Whisper + Claude)
- Compliance: 100% actions captured vs ~60% manually
Workflow: Zoom Webhook → Whisper Transcription → Claude Summary → Notion/Email
1. Webhook Node (Zoom)
- Event: recording.completed
- Download recording file
2. HTTP Request Node (Whisper API)
- URL: https://api.openai.com/v1/audio/transcriptions
- Method: POST (multipart/form-data)
- Body:
- file: {{$json.binary}}
- model: whisper-1
- language: en
- response_format: verbose_json (for timestamps)
3. HTTP Request Node (Claude API - Structuring)
Body:
{
"model": "claude-sonnet-4-5",
"max_tokens": 2000,
"system": "You are a note-taking assistant. Analyze the transcript and generate: executive summary (3 points), decisions made, assigned actions (person + deadline if mentioned), follow-up points, next steps. Markdown format.",
"messages": [
{
"role": "user",
"content": "Meeting transcript:\n\n{{$json.text}}\n\nParticipants: {{$json.meeting.participants}}"
}
]
}
4. Notion Node (Create Meeting Note)
- Database: Meetings
- Properties:
- Title: {{$json.meeting.topic}} - {{$json.meeting.start_time}}
- Participants: {{$json.meeting.participants}}
- Summary: {{$json.summary}}
- Decisions: {{$json.decisions}}
- Actions: {{$json.actions}}
- Full transcript: Toggle block
5. Email Node (Send to Participants)
Subject: 📝 Meeting notes: {{$json.meeting.topic}}
Hello,
Here's the automated summary of our meeting from {{$json.meeting.start_time}}:
## Executive Summary
{{$json.summary}}
## Decisions Made
{{$json.decisions}}
## Actions to Follow
{{$json.actions}}
## Next Steps
{{$json.next_steps}}
[View full notes in Notion]({{$json.notion_url}})
Workflow 6: Invoice Data Extraction (OCR + AI)
Use Case
An invoice PDF arrives by email or Drive. n8n uses OCR to extract text, Claude to identify supplier, tax-inclusive/exclusive amounts, date, number, billing lines, and automatically injects into your accounting software (Pennylane, QuickBooks) with consistency validation.
Estimated Savings
- Time: 10 min → 1 min per invoice
- Cost: €0.10 per invoice
- Error rate: 2% vs 8% in manual entry
Workflow: Email/Drive → OCR → Claude Extraction → Accounting Software
1. Email Trigger Node
- Subject contains: "invoice" OR "facture"
- Has attachment: PDF
2. Extract from File Node
- Mode: OCR (Tesseract)
- Language: eng+fra
3. HTTP Request Node (Claude API)
Body:
{
"model": "claude-sonnet-4-5",
"max_tokens": 1000,
"system": "You are an accounting expert. Extract structured data from this invoice: supplier (name, tax ID), number, issue date, due date, amount excl. tax, VAT, amount incl. tax, line items (description + quantity + unit price + total). Return valid JSON. If data is missing, use null.",
"messages": [
{
"role": "user",
"content": "Invoice OCR text:\n\n{{$json.text}}"
}
]
}
4. Function Node (Validate & Format)
const data = JSON.parse($input.item.json.content[0].text);
// Consistency validation
const calculatedTotal = data.amount_excl_tax * (1 + data.vat / 100);
const isValid = Math.abs(calculatedTotal - data.amount_incl_tax) < 0.5;
return {
json: {
...data,
validation: isValid ? 'OK' : 'REVIEW_NEEDED',
confidence: isValid ? 95 : 60
}
};
5. IF Node (Validation Check)
- If validation === 'OK' → Auto-import
- Else → Send to Slack for manual review
6. Pennylane/QuickBooks Node (Create Invoice)
- Supplier: {{$json.supplier.name}}
- Invoice number: {{$json.number}}
- Date: {{$json.issue_date}}
- Amount excl. tax: {{$json.amount_excl_tax}}
- VAT: {{$json.vat}}
- Total: {{$json.amount_incl_tax}}
- Line items: {{$json.lines}}
Workflow 7: Multi-Platform Social Media Content Generation
Use Case
You publish a blog article. n8n detects the publication (RSS/webhook), uses Claude to generate 5 adapted variations (Twitter thread, LinkedIn post, Instagram caption, Facebook post, newsletter teaser), schedules publication on each platform at the best time, and tracks performance.
Estimated Savings
- Time: 2h writing → 5 min validation
- Cost: €0.25 per article (Claude Sonnet)
- Reach: +300% thanks to optimized multi-distribution
Workflow: RSS/Webhook → Claude Content Gen → Schedule Posts
1. RSS Feed / Webhook Node
- URL: https://your-blog.com/feed
- Trigger: New item
2. HTTP Request Node (Claude API - Multi-Platform)
Body:
{
"model": "claude-sonnet-4-5",
"max_tokens": 2000,
"system": "You are a content marketing expert. Generate 5 variations of the same content adapted for: Twitter (thread max 5 tweets 280 chars), LinkedIn (professional post 1300 chars), Instagram (engaging caption + hashtags), Facebook (conversational post), Newsletter (150-word teaser). Keep tone and CTAs consistent. JSON format.",
"messages": [
{
"role": "user",
"content": "Article:\nTitle: {{$json.title}}\nSummary: {{$json.description}}\nURL: {{$json.link}}\n\nGenerate the 5 variations."
}
]
}
3. Function Node (Parse Variations)
const content = JSON.parse($input.item.json.content[0].text);
return [
{ json: { platform: 'twitter', content: content.twitter, url: $('RSS').item.json.link } },
{ json: { platform: 'linkedin', content: content.linkedin, url: $('RSS').item.json.link } },
{ json: { platform: 'instagram', content: content.instagram, url: $('RSS').item.json.link } },
{ json: { platform: 'facebook', content: content.facebook, url: $('RSS').item.json.link } },
{ json: { platform: 'newsletter', content: content.newsletter, url: $('RSS').item.json.link } }
];
4. Split In Batches Node
- Batch size: 1
5. Switch Node (Platform Router)
- twitter → Twitter API
- linkedin → LinkedIn API
- instagram → Meta Graph API
- facebook → Facebook Pages API
6. Twitter Node (Example)
- Authentication: OAuth
- Tweet text:
{{$json.content}}
🔗 Read the article: {{$json.url}}
7. Schedule Trigger Node (Best Time)
- Twitter: 9am, 1pm, 6pm
- LinkedIn: 8am, 12pm
- Instagram: 7pm, 9pm
Workflow 8: Automated Competitive Intelligence
Use Case
n8n daily scrapes 5 competitors' websites + their social media. Claude analyzes news (product launches, pricing, hiring), identifies threats/opportunities, generates a weekly report with strategic insights, and alerts on critical changes.
Estimated Savings
- Time: 5h/week manual monitoring → 30 min reading the report
- Cost: €2/week
- Reactivity: Real-time alerts vs late discovery
Workflow: Cron Daily → Scrape Competitors → Claude Analysis → Weekly Report
1. Schedule Trigger Node
- Interval: Every day at 6:00 AM
2. HTTP Request Node (Scrape - repeat for each competitor)
- URL: https://competitor-1.com/news
- Method: GET
- Extract: Cheerio selector (.news-section)
3. Merge Node (Combine all competitor data)
4. HTTP Request Node (Claude API - Analysis)
Body:
{
"model": "claude-sonnet-4-5",
"max_tokens": 3000,
"system": "You are a strategic analyst. Compare competitor news. Identify: price changes, new products, key hires, partnerships, positioning changes. Assess impact (critical/moderate/low) and suggest actions. Structured Markdown format.",
"messages": [
{
"role": "user",
"content": "Competitor data (last 7 days):\n\n{{$json.combined_data}}"
}
]
}
5. Airtable/Notion Node (Log Changes)
- Table: Competitive Intelligence
- Date: {{$now}}
- Competitor: {{$json.competitor_name}}
- Change: {{$json.change}}
- Impact: {{$json.impact}}
- Suggested action: {{$json.suggested_action}}
6. IF Node (Critical Alert)
- Condition: impact === "critical"
- True → Slack immediate alert
- False → Add to weekly digest
7. Schedule Node (Weekly Digest - Fridays 5pm)
- Aggregate all week's changes
- Generate executive summary with Claude
- Send email to leadership team
Workflow 9: Customer Feedback Analysis and Categorization
Use Case
Customer feedback arrives via NPS survey, app reviews, support tickets. n8n centralizes everything, uses Claude to detect sentiment, extract recurring themes (bugs, requested features, UX), score urgency, and generates a monthly dashboard with top 10 prioritized requests.
Estimated Savings
- Time: 10h/month manual analysis → 1h review
- Cost: €0.05 per feedback
- Product-market fit: Data-driven product decisions
Workflow: Multi-Source → Claude Analysis → Airtable Dashboard
1. Webhook Node (NPS Tool)
- Receives: score, comment, user_id
2. App Store Connect API Node
- Fetch: New reviews (rating + text)
3. Zendesk API Node
- Fetch: Closed tickets with customer_satisfaction
4. Merge Node (Combine all sources)
5. HTTP Request Node (Claude API - Analysis)
Body:
{
"model": "claude-haiku-4-5",
"max_tokens": 400,
"system": "Analyze this customer feedback. Extract: sentiment (positive/neutral/negative), themes (bug/feature/ux/pricing/support), urgency (high/medium/low), key_quote (most important sentence). JSON format.",
"messages": [
{
"role": "user",
"content": "Source: {{$json.source}}\nScore: {{$json.score}}\nComment: {{$json.comment}}"
}
]
}
6. Airtable Node (Create Record)
- Table: Feedback
- Date: {{$now}}
- Source: {{$json.source}}
- User: {{$json.user_id}}
- Sentiment: {{$json.sentiment}}
- Themes: {{$json.themes}} (multi-select)
- Urgency: {{$json.urgency}}
- Quote: {{$json.key_quote}}
7. Schedule Node (Monthly Aggregation)
- Trigger: First day of month
- Aggregate by themes
- Generate top 10 requested features
- Calculate sentiment evolution
8. HTTP Request Node (Claude - Executive Summary)
- Input: Aggregated stats
- Output: 1-page strategic report
- Send to: Product team + CEO
Workflow 10: Sales Pipeline Automation (Lead → Deal → Close)
Use Case
A lead enters your CRM. n8n orchestrates the entire pipeline: AI qualification score, data enrichment (LinkedIn, Clearbit), personalized email sequences generated by Claude based on profile, automatic follow-up, alerts to salespeople at the right time, deal stage updates, closing forecast.
Estimated Savings
- Conversion: +25% thanks to systematic follow-up
- Sales time: Focus on closing, not admin
- Cost: €0.50 per lead
Workflow: CRM Webhook → Enrichment → AI Emails → Stage Updates
1. HubSpot/Pipedrive Webhook Node
- Event: Contact created
- Stage: Lead
2. Clearbit Enrichment API Node
- Input: email
- Output: company_size, industry, technologies, funding
3. HTTP Request Node (Claude - Lead Scoring)
Body:
{
"model": "claude-sonnet-4-5",
"max_tokens": 500,
"system": "You are a B2B lead scoring expert. Analyze the data and assign a 0-100 score based on: fit (industry + size), intent (technologies used + budget), timing (growth). Justify the score and suggest sales approach (aggressive/nurture/disqualify).",
"messages": [
{
"role": "user",
"content": "Lead:\nName: {{$json.name}}\nCompany: {{$json.company}}\nIndustry: {{$json.industry}}\nSize: {{$json.company_size}}\nTechnologies: {{$json.technologies}}\nFunding: {{$json.funding}}"
}
]
}
4. IF Node (Score > 70 → Hot Lead)
5. HTTP Request Node (Claude - Personalized Email)
Body:
{
"model": "claude-sonnet-4-5",
"max_tokens": 600,
"system": "You are an expert SDR. Write a personalized prospecting email, hyper-targeted, that mentions a specific pain point for their industry and proposes concrete value. Tone: professional but not corporate. 150 words max.",
"messages": [
{
"role": "user",
"content": "Lead context:\n{{$json.context}}\n\nOur product solves: {{$env.VALUE_PROP}}\n\nWrite the email."
}
]
}
6. Gmail/SendGrid Node (Send Email)
- To: {{$json.email}}
- Subject: {{$json.custom_subject}}
- Body: {{$json.ai_email}}
7. Wait Node (3 days)
8. Check Email Response Node
- If replied → Update stage to "Engaged" + alert sales
- If not replied → Send follow-up (Claude-generated)
9. HubSpot Node (Update Deal)
- Score: {{$json.score}}
- Stage: Auto-updated based on engagement
- Forecast close date: {{$json.predicted_close}}
- Next action: {{$json.next_step}}
Production Best Practices for n8n + AI
1. Error Handling and Retry Logic
Node Settings (for each HTTP Request node):
- Continue on fail: true
- Retry on fail: 3 times
- Wait between retries: 5 seconds (exponential backoff)
Error Handling Node (after Claude API):
IF {{$json.error}} exists
→ Log to Sentry
→ Send alert to Slack
→ Fallback: Use default response or queue for manual review
2. AI Cost Optimization
- Cache repetitive prompts: use Redis to store identical AI responses
- Choose the right model: Haiku for classification ($0.25/M), Sonnet for generation ($3/M)
- Limit max_tokens: 300 tokens is enough for most classification tasks
- Batch processing: group similar requests into a single call with JSON array
3. Security and GDPR Compliance
# Environment variables (n8n)
ANTHROPIC_API_KEY=sk-ant-...
DATABASE_ENCRYPTION_KEY=...
WEBHOOK_SECRET=...
# EU-compliant hosting
- n8n: EU region (Frankfurt/Paris)
- Claude API: Data Processing Addendum signed
- PostgreSQL: encryption at-rest (AES-256)
# Logs and audit trail
- Enable n8n execution logs (30-day retention)
- Log all AI calls with anonymized user_id
- GDPR deletion workflow: purge data on user request
4. Monitoring and Alerts
# n8n Dashboard (Grafana + Prometheus)
Key metrics:
- Workflow execution time (p95 < 30s)
- AI API call latency (p95 < 3s)
- Error rate (< 2%)
- Cost per execution
Alerts:
- Error rate > 5% → PagerDuty
- AI cost > budget → Slack
- Workflow queue > 100 → Scale up
Comparison: n8n vs Zapier vs Make for AI
| Criterion | n8n | Zapier | Make |
|---|---|---|---|
| Cost (1000 executions/month) | €10 (VPS) + API costs | €100 + API costs | €40 + API costs |
| Claude API integration | ✅ Full control (HTTP node) | ⚠️ Via OpenAI Router (limited) | ✅ HTTP module |
| Complex system prompts | ✅ Unlimited | ❌ Max 500 characters | ✅ Flexible |
| Conversation history | ✅ PostgreSQL storage | ❌ Not supported | ⚠️ Via Data Store (limited) |
| Data residency (GDPR) | ✅ Self-hosted EU | ❌ US servers | ⚠️ EU option (extra cost) |
| Custom code (JS/Python) | ✅ Unlimited | ⚠️ Limited (Code by Zapier) | ✅ Full support |
| Best for | Production AI, dev teams | No-code teams, prototypes | Middle ground |
Resources and Training
To master n8n and production AI automation, explore our AI training courses covering Claude API integrations, complex workflow design, and production patterns. Our courses are OPCO eligible in France (potential out-of-pocket cost: €0).
We also cover advanced prompts and multi-model orchestration in our training programs.
Frequently Asked Questions
Is n8n really free for AI automation?
Yes. n8n is open-source and can be self-hosted for free (Docker, VPS). Only AI APIs (Claude, OpenAI, Whisper) are billed per usage. A self-hosted AI workflow typically costs €5-50/month depending on volume, versus €200-500/month for equivalent Zapier + Make.
What's the difference between n8n and Zapier for AI?
n8n is open-source, self-hosted, and offers full data control. Zapier is proprietary SaaS with per-task billing. For AI: n8n allows complex prompts and conversation history management, Zapier limits to simple prompts. Cost: n8n ~€10/month + API calls, Zapier ~€100/month + API calls.
Can I use Claude API with n8n?
Yes. n8n provides an HTTP Request node that allows calling Anthropic's Claude API. You can also use the OpenAI Router node compatible with Claude. The workflows in this article use Claude 4.5 Sonnet for generation and Haiku for fast classification.
How to deploy an n8n workflow to production?
Three options: (1) Docker Compose on a VPS (€5/month at Hetzner or DigitalOcean), (2) n8n Cloud (managed deployment, from €20/month), (3) Kubernetes to scale to thousands of executions/day. To start: Docker Compose is sufficient and can handle 10k workflows/month.
Can n8n workflows process sensitive data?
Yes, it's even recommended vs Zapier. With self-hosting, data never leaves your infrastructure. For GDPR compliance: (1) host n8n in EU, (2) choose GDPR-compliant AI APIs (Claude API is GDPR-compliant), (3) encrypt PostgreSQL database, (4) enable audit logs.