From atum-nocode
n8n workflow automation pattern library — leverages the n8n-mcp server (declared in this plugin's .mcp.json) which exposes 1396 n8n nodes (812 core + 584 community) for AI agents to build production-ready workflows. Covers core n8n concepts (workflows, executions, triggers, regular vs trigger nodes, expressions with $node syntax), trigger types (webhook for HTTP, cron for scheduled, polling for periodic checks, manual, event-based for SaaS apps), data flow (item-based processing, pinning data for testing, JMESPath/JSONPath in expressions), error handling (Error Workflow, retry on fail, continue on fail, conditional branching), credentials management (encrypted at rest, OAuth 2.0 flows, API key headers), self-hosted vs n8n.cloud trade-offs (cost, scalability, sovereignty), Code node for custom JavaScript / Python, sub-workflows for reusability, version control via Git (export workflows as JSON), CI/CD with n8n CLI, monitoring and execution logs, AI agent nodes (LangChain integration, OpenAI / Anthropic / Ollama nodes), and migration from Zapier or Make. Use when building automation workflows with n8n, choosing between Zapier / Make / n8n, debugging complex workflows, or migrating from another platform. Mentions the official `n8n` MCP server — Claude Code can directly query, create, and manage n8n workflows at runtime.
npx claudepluginhub arnwaldn/atum-plugins-collection --plugin atum-nocodeThis skill uses the workspace's default tool permissions.
Patterns canoniques pour utiliser **n8n** (open source self-hosted ou n8n.cloud) en s'appuyant sur le **MCP server n8n-mcp** déclaré dans `plugins/atum-nocode/.mcp.json`.
Executes pre-written implementation plans: critically reviews, follows bite-sized steps exactly, runs verifications, tracks progress with checkpoints, uses git worktrees, stops on blockers.
Guides idea refinement into designs: explores context, asks questions one-by-one, proposes approaches, presents sections for approval, writes/review specs before coding.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
Patterns canoniques pour utiliser n8n (open source self-hosted ou n8n.cloud) en s'appuyant sur le MCP server n8n-mcp déclaré dans plugins/atum-nocode/.mcp.json.
MCP server n8n disponible : exposes 1396 nodes (812 core + 584 community). Claude Code peut directement créer / lister / inspecter / debug des workflows n8n.
Prérequis utilisateur : Node.js + instance n8n accessible (self-hosted Docker ou n8n.cloud) + API key.
# Docker
docker run -d --name n8n -p 5678:5678 \
-v n8n_data:/home/node/.n8n \
-e N8N_BASIC_AUTH_ACTIVE=true \
-e N8N_BASIC_AUTH_USER=admin \
-e N8N_BASIC_AUTH_PASSWORD=$(openssl rand -base64 24) \
-e N8N_HOST=n8n.example.com \
-e WEBHOOK_URL=https://n8n.example.com/ \
-e N8N_PROTOCOL=https \
n8nio/n8n
Génère ensuite l'API key dans Settings → API → Create API Key.
# Set env vars pour le MCP
export N8N_API_URL=https://n8n.example.com/api/v1
export N8N_API_KEY=your-api-key
Trigger
├── Webhook (HTTP endpoint)
├── Cron (scheduled)
├── Manual (test)
├── Polling (every N min, check API for changes)
└── Event (Slack mention, GitHub PR opened, Stripe payment, etc.)
→
Regular nodes (process data)
├── HTTP Request (REST APIs)
├── Code (JS/Python custom)
├── IF / Switch (branching)
├── Set / Merge / Split (data transform)
├── 1396 SaaS integrations (Slack, Notion, Airtable, Stripe, ...)
└── AI nodes (OpenAI, Anthropic, LangChain)
→
Output / action
├── HTTP Request (POST somewhere)
├── Send email
├── Slack message
├── Create row in DB / Airtable / Notion
└── Trigger another workflow
[Webhook trigger]
│
▼
[Validate input (Code or IF)]
│ ✅
▼
[Process (HTTP, DB, transform)]
│
▼
[Webhook response: 200 OK]
// Code node — validation
const body = $input.first().json
if (!body.email || !body.email.includes('@')) {
return [{ json: { error: 'Invalid email' }, statusCode: 400 }]
}
return [{ json: body }]
[Main workflow]
│
├──> [Execute Sub-Workflow: send_notification(channel, message)]
│
└──> [Execute Sub-Workflow: log_event(event_type, payload)]
Sub-workflows = bibliothèque interne de "fonctions" réutilisables. Importer le JSON du sub-workflow et l'appeler avec Execute Sub-Workflow node.
[Trigger]
│
▼
[HTTP Request (max 3 retries, 5s wait)]
│
├── ✅ Success → [Continue]
└── ❌ Fail → [Error Workflow: notify Slack + write to error DB]
Activer dans le node Settings → "Continue On Fail" + "Retry On Fail (max 3, wait 5s)".
Configurer un Error Workflow global dans Settings → Workflow → Error Workflow → sélectionner le workflow d'erreur.
Règle : ne JAMAIS hardcoder un token dans une expression {{$json.api_key}}. Toujours via un Credential node.
// JavaScript
const items = $input.all()
const enriched = items.map(item => {
return {
json: {
...item.json,
timestamp: new Date().toISOString(),
processed: true,
},
}
})
return enriched
# Python (via Code node Python mode)
import json
from datetime import datetime
items = _input.all()
enriched = []
for item in items:
item_data = item.json
item_data['timestamp'] = datetime.utcnow().isoformat()
item_data['processed'] = True
enriched.append({'json': item_data})
return enriched
[Webhook : user message]
│
▼
[OpenAI Chat Model: classify intent]
│
├── intent="support" ──> [Create Zendesk ticket]
├── intent="sales" ────> [Create HubSpot deal]
└── intent="other" ────> [Send to Slack #general]
n8n a des nodes natifs OpenAI, Anthropic, Ollama (self-hosted), ainsi que LangChain agents pour des workflows agentic plus complexes.
n8n stocke les workflows dans sa DB. Pour Git :
# Export tous les workflows en JSON
n8n export:workflow --all --output=./workflows/
# Import depuis JSON
n8n import:workflow --separate --input=./workflows/
Workflow Git :
n8n import:workflow sur l'instance prod| Critère | Zapier | Make (Integromat) | n8n |
|---|---|---|---|
| Open source | ❌ | ❌ | ✅ |
| Self-hostable | ❌ | ❌ | ✅ |
| Prix | $$$ par task | $$ par operation | gratuit self-host / $$ cloud |
| Nodes / apps | 6000+ | 1500+ | 1396 |
| Custom code | Limité (JS) | Limité (JSON) | JS + Python complet |
| Workflows complexes | Linéaire | Visuel branching | Visuel branching avancé |
| AI nodes natifs | Limité | Limité | LangChain complet |
| Sovereignty / GDPR | US-hosted | EU possible | Self-host total |
| Migration ATUM | À éviter | OK | Recommandé |
Verdict 2026 : pour les agences avec besoins GDPR ou cost control, n8n self-hosted est le meilleur choix. Pour les non-tech, n8n.cloud reste utilisable.
make-scenario-builder (ce plugin)airtable-schema-manager (ce plugin)notion-automation (ce plugin)no-code-automation-expert (ce plugin)api-designer (atum-stack-backend)