Deploy n8n 3.0
n8n 3.0 with workers & SearXNG web search. n8n v3 Pro Stack. n8n3/n8nv3
SearXNG
Just deployed
n8n-worker
Just deployed
Just deployed
/var/lib/postgresql/data
n8n-main
Just deployed
Redis
Just deployed
/bitnami
Code Runners
Just deployed
Deploy n8n 3.0 / n8n v3 Pro Stack with SearXNG on Railway
For n8n 3.0 on Railway with queued executions, external JavaScript/Python runners and private web search, use n8n v3 Pro Stack with SearXNG. It pins n8n 3.0.0 and matching runners, with PostgreSQL, Redis and database-backed file storage across six services. Deploy without an AI key, then configure your integrations. This creates a fresh V3 installation; review the pre-release and migration notes before moving existing workflows.
No AI API key, external database account, S3 account, or search API subscription is required to deploy. Add provider credentials inside n8n for the integrations you use. Railway hosting and any external model or sandbox usage are billed separately.
Version note: 3.0.0 was marked pre-release when this stack was validated on October 30, 2026. The template pins that version and matching runners. Check the official release before using it for critical workflows. This is a fresh installation, not an automatic upgrade of your n8n 2.0 deployment.
What this Pro Stack includes
| Service | Role |
|---|---|
| n8n-main | Editor, authentication, schedules, HTTPS webhooks |
| n8n-worker | Queued workflow execution; concurrency 5 |
| Code Runners | External JavaScript and native Python Code nodes |
| Postgres | Accounts, workflows, encrypted credentials, executions, binary files |
| Redis | Persistent execution queue |
| SearXNG | Private web search with JSON responses enabled |
Only the editor has a public web domain. Worker and runner health checks are separate. Standard Railway PostgreSQL and Redis include their usual variables, generated passwords, private hostnames, and volumes. This is a single-worker stack, not a high-availability cluster.
Why use this instead of a single n8n container?
- Integrated private search: SearXNG is connected to Assistant/Agents on both editor and worker, and is also usable from ordinary workflows.
- Separate execution capacity: production webhook and manual runs use queue mode. The worker's start command controls concurrency.
- Isolated code runners: JavaScript and Python run in a separate service with an authenticated private broker connection. It receives its broker token, not your database password or encryption key. Idle language processes stop after 15 seconds and restart for new work.
- Shared file storage: editor and worker use PostgreSQL for binary data, avoiding separate container filesystems. No shared filesystem, S3 bucket, or n8n external-storage licence is needed for this configuration. Files are limited to 64 MiB by default.
- Persistent encrypted credentials: editor and worker share a generated master encryption key through a variable reference. Core data lives in PostgreSQL; Redis persists queue state.
- Controlled updates and retention: matching n8n and runner images are pinned by digest. Completed execution data is pruned after seven days, with at most 10,000 records retained.
A single container can cost less for light workloads. “Pro Stack” describes the infrastructure, not paid n8n Business or Enterprise features.
First deployment
- Deploy into your Railway workspace; no AI provider key is required to start.
- Wait for all six services to become healthy. Initial database migrations may take a few minutes.
- Open the
n8n-maindomain and promptly create your owner account in n8n's normal setup screen. - Create and publish a workflow. Production webhook URLs already use your HTTPS domain.
- Add credentials inside n8n for OpenAI, Anthropic, other model providers, or API integrations as needed.
- For Assistant web search, select SearXNG and use the configuration below if the setup screen asks for a URL.
The initial timezone is Europe/Paris. Change GENERIC_TIMEZONE and TZ consistently on editor and worker for another timezone.
Included SearXNG web search
The template supplies N8N_INSTANCE_AI_SEARXNG_URL on editor and worker. Select SearXNG in Assistant setup. If asked for a URL, copy that variable's value from your deployed n8n-main service, for example http://searxng.railway.internal:8080. A search credential saved inside n8n takes precedence over this environment default.
SearXNG has no public domain, does not reuse the execution queue's Redis, and recreates its declarative settings from Railway variables at startup. Its ephemeral cache needs no volume. There is no search API subscription, but Railway charges for the container. Upstream search engines can rate-limit cloud IPs or return incomplete results.
Search from a regular workflow
Add an HTTP Request node:
- Method:
GET - URL:
http://searxng.railway.internal:8080/search - Query parameters:
q=your search termsandformat=json
If you renamed the service, use the actual hostname from N8N_INSTANCE_AI_SEARXNG_URL. The response contains a results array with titles, links, and snippets. This private address is accessible from n8n inside Railway, not from your browser.
Use it for research, documentation lookup, or AI workflows with your own model credentials.
Assistant code sandbox is separate
Code Runners are not the Assistant code sandbox. Port 5679 on the worker is its task broker, not a sandbox endpoint. n8n's bundled sandbox requires privileged Docker-in-Docker, which Railway services do not provide.
For sandbox-backed Assistant features, configure Daytona with your own API key, or host the official n8n sandbox stack on an external compatible Docker host. Configure your model provider separately. Ordinary workflows and JavaScript/Python Code nodes do not require that extra sandbox.
Code, files, and additional workers
The external runners keep module allowlists restricted. Additional Python or npm dependencies require a customised runner image and explicit allowlists; environment variables alone do not install packages. The task timeout is set to 300 seconds on n8n and the runners.
Both n8n services use N8N_DEFAULT_BINARY_DATA_MODE=database. Keep this consistent across workers. Large files increase database storage, memory, and backup costs; review the 64 MiB limit before raising it.
Each additional worker needs its own runner connected to that worker's private broker. Copy the database, queue, encryption-key, and binary-storage references.
Persistence, backups, and maintenance
PostgreSQL and Redis have persistent volumes. n8n containers are replaceable: core application data and binary files live in PostgreSQL, with encryption material supplied through Railway variables. Manually added files inside application containers are not durable.
Enable PostgreSQL volume backups in Railway and test a restore. A persistent volume alone is not a backup. Preserve N8N_ENCRYPTION_KEY securely: a database backup without the matching key cannot recover encrypted credentials.
Keep editor, worker, and databases running for scheduled jobs and webhooks. Language runners idle internally; this stack does not rely on Railway serverless sleep. All six services and their storage contribute to hosting costs.
For planned maintenance, stop new requests at the editor, drain and stop the worker, then stop runners and databases. Explicit runner shutdown deadlines let the broker drain cleanly. Let queued work finish, or test your interruption and retry procedure with a disposable workflow.
Updates and migration from n8n 2.0
Back up the database and encryption key, then update editor, worker, and runners together to a matching release. Digest pinning prevents unreviewed upgrades through a moving latest tag.
For existing workflows, review the v3 breaking changes, use n8n's migration report, and test copies first. This template creates a fresh database; it does not migrate your old installation. Imported credential backups need a compatible encryption-key procedure.
Frequently asked questions
Can I deploy n8n v3 without an AI API key?
Yes. The editor, execution queue, code runners, SearXNG, and non-AI workflows start without one. Configure provider credentials in n8n when needed.
Does Pro Stack include paid features or unrestricted community nodes?
No. Follow n8n's supported installation and licensing process. v3 disables unverified community packages by default; paid features still require the appropriate licence.
Recorded validation
Validation on October 30, 2026 covered queue execution, JavaScript/Python Code nodes, worker use of encrypted credentials, protected downloads, database persistence, and container replacement. SearXNG returned JSON results from a workflow before and after container replacement. These checks do not guarantee every third-party node, model, or search engine.
Official references
- Queue mode
- External task runners
- Binary data configuration
- SearXNG search API
- n8n Assistant configuration
Other deployment choices from this publisher
Template Content
SearXNG
ghcr.io/searxng/searxngn8n-worker
n8nio/n8n:3.0.0n8n-main
n8nio/n8n:3.0.0Redis
railwayapp/redisCode Runners
n8nio/runners:3.0.0