Skip to content

For n8n

n8n on Edge, with no execution meter

The n8n editor, webhooks and workers on your own VMs. Start on one machine, switch to queue mode with Redis and workers as volume grows, and keep every API credential on your network.

Free to sign up. No egress fees on any product.

Why n8n on Edge

Automation at any volume, without a meter running

n8n does what Zapier and Make do. On Edge it runs on your VMs, talks to your private network and costs the same whether a workflow fires ten times or a million.
  • Hundreds of integrations

    Webhooks, schedules, HTTP requests and AI nodes, plus hundreds of built-in integrations. All on your VM.
  • Queue mode when you need it

    Start with one process. Set EXECUTIONS_MODE=queue and Redis hands jobs to as many workers as you run.
  • Credentials stay on your network

    Stripe, OpenAI, Salesforce and Slack keys are encrypted with your N8N_ENCRYPTION_KEY and stored in your own Postgres.
  • Webhooks behind the CDN

    Inbound webhooks arrive over TLS at the CDN, with DDoS protection included, and pass through uncached to the VM.
  • Private network for workers

    Worker VMs reach Redis and Postgres over a free private network in the same region. Nothing internal faces the internet.
  • No per-execution bills

    n8n Cloud prices by executions. On Edge, a million runs a month costs the same as a thousand.

Reference architecture

How n8n maps to Edge

Queue mode splits n8n in two. The main process serves the editor and receives triggers; workers run the executions. Redis sits between them and Postgres holds everything.
  • Compute

    Runs the n8n main process, workers, Redis and Postgres. One VM to start, more over a private network later.

  • Storage

    Nightly pg_dump and workflow exports. Binary data can live here too with n8n's S3 mode (Enterprise licence).

  • CDN

    TLS and DDoS protection for the editor and inbound webhooks, with caching bypassed on /webhook/*.

  • DNS

    Anycast DNS for n8n.acme.com.

  • Shield

    Scores public form submissions before they trigger a workflow.

Deploy

n8n in queue mode, in five steps

The official image, Postgres and Redis. Start everything on one VM, then move workers out when you need to.
  1. 01

    Create a VM and a backup bucket

    A Performance VM (4 vCPU, 8 GiB) runs the main process, two workers, Redis and Postgres side by side.

    shell
    $ edge compute create --name n8n --size s-4vcpu-8gb \
        --image ubuntu-24 --region london --script docker
    $ edge storage create n8n-backups
  2. 02

    Configure n8n for queue mode

    Every n8n process, main and workers, needs the same encryption key. Keep it safe: it decrypts your credentials.

    .env
    N8N_HOST=n8n.acme.com
    N8N_PROTOCOL=https
    WEBHOOK_URL=https://n8n.acme.com/
    N8N_ENCRYPTION_KEY=<openssl rand -hex 32>
    GENERIC_TIMEZONE=Europe/London
    
    DB_TYPE=postgresdb
    DB_POSTGRESDB_HOST=postgres
    DB_POSTGRESDB_DATABASE=n8n
    DB_POSTGRESDB_USER=n8n
    DB_POSTGRESDB_PASSWORD=<password>
    
    EXECUTIONS_MODE=queue
    QUEUE_BULL_REDIS_HOST=redis
    OFFLOAD_MANUAL_EXECUTIONS_TO_WORKERS=true
  3. 03

    Write the compose file

    Main and workers share one definition. The worker service runs the same image with the worker command.

    docker-compose.yml
    x-n8n: &n8n
      image: docker.n8n.io/n8nio/n8n
      env_file: .env
      restart: always
      depends_on: [postgres, redis]
    
    services:
      n8n:
        <<: *n8n
        ports: ["127.0.0.1:5678:5678"]
        volumes: ["n8n-data:/home/node/.n8n"]
      n8n-worker:
        <<: *n8n
        command: worker --concurrency=10
      redis:
        image: redis:7-alpine
      postgres:
        image: postgres:16
        environment:
          POSTGRES_USER: n8n
          POSTGRES_PASSWORD: ${DB_POSTGRESDB_PASSWORD}
          POSTGRES_DB: n8n
        volumes: ["pg-data:/var/lib/postgresql/data"]
    
    volumes:
      n8n-data:
      pg-data:
  4. 04

    Start it and put the CDN in front

    Add a bypassCache rule for /webhook/* and /rest/* in the deployment's configuration so nothing dynamic is ever cached.

    shell
    $ docker compose up -d --scale n8n-worker=2
    $ edge cdn create --name n8n
    $ edge cdn domains add cdn-a1b2c3 \
        --domain n8n.acme.com \
        --origin https://<vm-ip>
  5. 05

    Scale workers out when you need to

    Workers only need Redis and Postgres. Publish both on the private interface only, and point new worker VMs at the main VM's private IP.

    shell
    $ for i in 1 2 3; do
        edge compute create --name n8n-worker-$i --size s-2vcpu-4gb \
          --image ubuntu-24 --region london --script docker
      done
    
    # attach each VM to the private network in the console, then set
    #   QUEUE_BULL_REDIS_HOST=10.185.0.2
    #   DB_POSTGRESDB_HOST=10.185.0.2
    $ docker compose up -d n8n-worker

Prefer to hand it off? Give the job to your AI agent or have our engineers do it.

What it costs

Half a million executions, one VM

Hosted automation tools price every run. Here the price is the VM, whatever your workflows do.
  • Unlimited executions and active workflows
  • Add Standard worker VMs at $19.24/mo each
  • Private networking between VMs is free
  • Zero egress on API calls, webhooks and exports
See compute pricing

Estimated monthly bill on Edge

~500k executions a month · queue mode on one VM

USD
  • Compute$38.47Performance VM · 4 vCPU · 8 GiB · 160 GB NVMe
  • CDN$0.20~1.3M requests (webhooks + editor), first 500k free, then $0.25/M
  • Storage$0.1515 GB of backups and exports, first 5 GB free
  • DNS$0.00Zone and records for n8n.acme.com
  • Egress$0.00
Total$38.82

FAQ

n8n on Edge, answered

Something else? Ask an engineer.
How does this compare to n8n Cloud or Zapier?
n8n Cloud prices by executions and Zapier by tasks. Self-hosted on Edge is a flat VM price. The trade-off is a VM to maintain, which the unlimited executions and on-network credentials usually justify.
When should I switch to queue mode?
Regular mode runs everything in one process and handles plenty of workflows. Move to queue mode when long executions start delaying others, or when you want to add workers without touching the editor.
Can binary data go to Edge Storage?
n8n's S3 binary-data mode works with any S3-compatible endpoint, including https://storage.edge.network, but it's an Enterprise-licensed feature. On the Community Edition, back up Postgres and exports to Edge Storage instead.
How do I back up workflows?
Workflows and encrypted credentials live in Postgres, so a nightly pg_dump to Edge Storage covers them. n8n export:workflow --all adds portable JSON copies alongside.
Can I use AI nodes with my own models?
Yes. n8n has nodes for OpenAI-compatible APIs and Ollama. Run Ollama or llama.cpp on an Edge CPU VM for small models and keep the whole pipeline on your network.

Run n8n on your terms

Start free and move your Zapier or Make workflows onto VMs you control.

Free tiers hard-cap. Nothing bills until you add a card.