n8n¶

What it is¶

n8n is an extendable, source-available workflow automation platform with a visual node editor, API integrations, and first-class support for AI-powered workflow steps.

What problem it solves¶

It replaces repetitive manual operations across tools and teams. Unlike cloud-only automation products, it can be self-hosted, so workflow logic, execution history, and sensitive data stay in your environment.

Where it fits in the stack¶

Automation & Orchestration. It is the control plane for cross-tool business processes.

Typical use cases¶

Autonomous document/email operations: classify incoming content, extract fields, route to owners.
Cross-system process automation: connect email, CRM/ERP, ticketing, chat, storage, and databases.
AI-assisted operations: triage, summarize, draft responses, and escalate only low-confidence cases.

Strengths¶

Visual + programmable: easy to build visually, still supports advanced expressions/logic.
Self-hostable: private automation with infrastructure you control.
Broad integrations: large ecosystem of built-in and community nodes.
Good ops model: retries, execution logs, error workflows, queue mode.

Limitations¶

Learning curve: robust flows require strong data modeling and error handling.
Scale design required: business-critical usage needs queue mode, persistent DB, and observability.
Credential discipline: secrets management and environment separation must be done explicitly.

When to use it¶

You want long-running, auditable business automations.
You need privacy and self-hosting.
You want AI-assisted processes with clear human-approval boundaries.

When not to use it¶

For one-off scripts or tiny automations with no lifecycle.
When you do not want to own operations/security for automation infrastructure.

Licensing and cost¶

Source Available: Yes (Fair-code license)
Cost: Free (Self-hosted) / Paid (Cloud/Enterprise)
Self-hostable: Yes

Getting started¶

Installation (Docker Compose, production-oriented baseline)¶

services:
  postgres:
    image: postgres:16
    environment:
      - POSTGRES_USER=n8n
      - POSTGRES_PASSWORD=${POSTGRES_PASSWORD}
      - POSTGRES_DB=n8n
    volumes:
      - ./postgres_data:/var/lib/postgresql/data

  n8n:
    image: docker.n8n.io/n8nio/n8n:latest
    container_name: n8n
    depends_on:
      - postgres
    ports:
      - 5678:5678
    volumes:
      - ./n8n_data:/home/node/.n8n
      - ./n8n_files:/files
    environment:
      - DB_TYPE=postgresdb
      - DB_POSTGRESDB_HOST=postgres
      - DB_POSTGRESDB_PORT=5432
      - DB_POSTGRESDB_DATABASE=n8n
      - DB_POSTGRESDB_USER=n8n
      - DB_POSTGRESDB_PASSWORD=${POSTGRES_PASSWORD}
      - N8N_HOST=n8n.local
      - N8N_PORT=5678
      - N8N_PROTOCOL=https
      - N8N_ENCRYPTION_KEY=${N8N_ENCRYPTION_KEY}
      - NODE_ENV=production
    restart: unless-stopped

Project definitions: JSON workflows and YAML infrastructure¶

If you want to version-control an n8n project:

Workflow definitions: native n8n format is JSON.
Infrastructure/deployment: usually YAML (docker-compose.yml, CI workflows, Kubernetes manifests).
Repository manifest (optional): your own YAML index for team governance can map business processes to workflow JSON files.

Example repository layout:

automation-n8n/
  docker-compose.yml
  .env.example
  n8n/
    workflows/
      010-email-intake.json
      020-quote-draft.json
      030-shipment-updates.json
    workflow-manifest.yaml
  docs/
    runbook.md

Example workflow-manifest.yaml (repo convention, not native n8n format):

project: wine-import-export-ops
owner: operations
workflows:
  - name: email-intake-and-triage
    file: n8n/workflows/010-email-intake.json
    criticality: high
  - name: quote-draft-and-follow-up
    file: n8n/workflows/020-quote-draft.json
    criticality: high

CLI examples¶

# Export all workflows as separate JSON files
docker compose exec n8n n8n export:workflow --all --separate --output=/files/workflows

# Import all workflows from a folder
docker compose exec n8n n8n import:workflow --separate --input=/files/workflows

# Export credentials for backup/migration (handle securely)
docker compose exec n8n n8n export:credentials --all --output=/files/credentials.json

How to use AI to run n8n operations (not only AI nodes inside workflows)¶

Use AI in three roles around n8n:

Designer role: convert SOPs into workflow specs, node maps, and test cases.
Operator role: review failed executions, classify root cause, propose retries/fixes.
Optimizer role: identify manual bottlenecks and propose new automations weekly.

Guardrails:

AI can propose/create/update workflows and documentation.
AI cannot approve legal commitments, banking actions, or final contractual terms.
High-risk actions require human approval nodes and explicit audit logs.

Example: Wine import/export email automation program¶

Goal: n8n handles most email operations end-to-end while humans keep control of legal and banking decisions.

Core workflows:

Inbound email triage
Trigger: IMAP/Gmail new email.
Steps: classify (supplier, customer, logistics, compliance, finance), extract entities (product, quantity, Incoterm, destination, ETA).
Output: route to CRM/ERP/ticket queue and attach suggested response draft.
Quote request automation
Trigger: inbound RFQ emails.
Steps: fetch current catalog/pricing rules, draft quote reply, create follow-up tasks, schedule reminder if no response.
Output: ready-to-send draft + SLA timer.
Shipment status and exception handling
Trigger: forwarder/carrier updates by email/API.
Steps: parse status, update shipment record, notify customer on milestones, escalate delays/anomalies.
Output: near-real-time customer comms without manual copy/paste.
Collections and accounts comms (non-payment execution)
Trigger: invoice due/overdue event.
Steps: send reminders, track response sentiment, escalate disputed cases.
Output: n8n drives communication cadence; finance team approves/executes payment actions externally.
Compliance document orchestration
Trigger: missing/updated document request.
Steps: request docs, validate required fields, route to legal/compliance for approval.
Output: complete compliance packet ready for human sign-off.

Human-only boundaries (explicit)¶

Bank transfers, payment approval, and settlement.
Legal review/approval of contracts and regulatory commitments.
Any action above predefined risk threshold.

Minimal AI prompt contract for email triage¶

Classify this email into one of:
supplier, customer, logistics, compliance, finance, spam.

Return strict JSON:
{
  "category": "...",
  "urgency": "low|medium|high",
  "entities": {
    "counterparty": "",
    "product": "",
    "quantity": "",
    "destination": "",
    "incoterm": ""
  },
  "needs_human_approval": true|false,
  "reason": ""
}

How to keep this n8n capability expanding over time (Jules loop)¶

Use a recurring Jules task focused on n8n value growth:

Pull top failed/slow executions from last 24h.
Propose one additive workflow improvement.
Add one new test case for that workflow.
Open a PR with workflow JSON + docs update + rollback note.

This keeps n8n improving as an operating system for your business processes, not just as isolated automations.

Useful AI-adjacent integrations¶

Tavily is a good fit when workflows need web search or research enrichment before summarization or routing.
Playwright is useful when the target system lacks a stable API and a browser automation fallback is acceptable.
Supabase works well as lightweight state storage for workflow memory, queue snapshots, or approval state outside n8n itself.
Replicate and ElevenLabs are useful when automations need media generation rather than text-only inference.

Company starter examples¶

Lead intake system: web form or email -> enrich with Tavily -> store in Supabase -> create Workspace assets -> create follow-up tasks.
Sales prep workflow: account name -> research via DeerFlow/Tavily -> save brief -> send summary to Workspace docs or CRM.
Content production loop: topic queue -> research -> generate draft -> human review -> publish checklist -> performance tracking.

Selection comments¶

n8n should usually be the control plane, not the only intelligence layer.
Let specialized tools handle search, memory, browser automation, and inference; let n8n coordinate timing, retries, approvals, and auditability.

Ollama (Use as AI backend)
Zapier (Alternative)
Make (Alternative)
OpenRouter (Multi-provider LLM routing for prompts/models)
Home Assistant (Example of broader automation ecosystem integration)

n8n¶

What it is¶

What problem it solves¶

Where it fits in the stack¶

Typical use cases¶

Strengths¶

Limitations¶

When to use it¶

When not to use it¶

Licensing and cost¶

Getting started¶

Installation (Docker Compose, production-oriented baseline)¶

Project definitions: JSON workflows and YAML infrastructure¶

CLI examples¶

How to use AI to run n8n operations (not only AI nodes inside workflows)¶

Example: Wine import/export email automation program¶

Human-only boundaries (explicit)¶

Minimal AI prompt contract for email triage¶

How to keep this n8n capability expanding over time (Jules loop)¶

Useful AI-adjacent integrations¶

Company starter examples¶

Selection comments¶

Backlog¶

Sources / References¶

Contribution Metadata¶

n8n¶

What it is¶

What problem it solves¶

Where it fits in the stack¶

Typical use cases¶

Strengths¶

Limitations¶

When to use it¶

When not to use it¶

Licensing and cost¶

Getting started¶

Installation (Docker Compose, production-oriented baseline)¶

Project definitions: JSON workflows and YAML infrastructure¶

CLI examples¶

How to use AI to run n8n operations (not only AI nodes inside workflows)¶

Example: Wine import/export email automation program¶

Human-only boundaries (explicit)¶

Minimal AI prompt contract for email triage¶

How to keep this n8n capability expanding over time (Jules loop)¶

Useful AI-adjacent integrations¶

Company starter examples¶

Selection comments¶

Related tools / concepts¶

Backlog¶

Sources / References¶

Contribution Metadata¶