Railway is a usage-billed container platform with a polished canvas, zero-config builds and one-click database templates, running on its own hardware. Quave ONE runs your apps and managed databases on our regions or in your own cloud account.
Facts checked on November 10, 2026, with a source link on every row.
Railway is great for getting something running fast: per-second billing on what you actually use, automatic PR environments, Dockerfile and zero-config builds, and SOC 2 Type II. For side projects and small teams that live in the dashboard, it is a strong choice.
Quave ONE fits teams that want support from people by email, a São Paulo region, CPU- and memory-based autoscaling, and the option to run the same platform in their own cloud account. Railway support is community or forum-based below Enterprise and it has no region in South America.
| Quave ONE | Railway | |
|---|---|---|
| Pricing model | Direct: usage-based, $7.50 per zCloud per month (0.5 vCPU + 0.5 GB RAM), billed hourly. Overage lines apply to disk above 5 GB per zCloud, backups above 30 GB per environment, and builds above 600 minutes per month. Connect: a platform license from $399/month, and you pay your cloud provider directly.Source | Plan fee credited toward usage (Hobby $5, Pro $20 per month), then per-second billing on actual use: $20 per vCPU-month and $10 per GB RAM-month. Builds are free.Source |
| Smallest always-on app | 1 zCloud (0.5 vCPU, 0.5 GB RAM) for $7.50/month in standard regions (US-5, EU-3, BR-1), after a free Professional upgrade review. Logs, metrics, alerts, backups, MCP and support are included. EU-4 London is $26.50.Source | Billed by real usage, not allocation: an app averaging 0.5 vCPU and 0.5 GB RAM all month uses about $15 of compute; an idle app uses less.Source |
| Free option | Hobby account: deploy apps and databases to the US-1 Hobby region to evaluate. No charge and no credit card. Not for production.Source | 30-day trial with a one-time $5 credit, no credit card. Then a Free plan with $1 of credit per month, capped at 1 vCPU / 0.5 GB per service.Source |
| Bandwidth | No extra network fees. In US-5, BR-1, EU-3 and EU-4, 10 GB of outbound traffic per zCloud per month is included; contact support if you need more.Source | Egress is $0.05 per GB for services.Source |
| Managed databases | Managed MongoDB, PostgreSQL, MySQL, Redis, RabbitMQ and ClickHouse, with managed backups kept 7 days by default. A single-node MongoDB starts at 1 zCloud ($7.50/month).Source | Railway-maintained templates for PostgreSQL, MySQL, Redis and MongoDB, with HA variants. No published database uptime SLA.Source |
| Long-running apps and WebSockets | Long-running containers, so WebSockets work. Idle connections close after 60 seconds without data by default; keep-alive pings (such as Meteor DDP heartbeats) keep them open.Source | Always-on containers; WebSockets have no duration or idle limit at the edge. No sticky sessions across replicas.Source |
| Regions | Direct: Ashburn (US-5), Frankfurt (EU-3), London (EU-4) and São Paulo (BR-1). Connect: your own cloud account; the license covers one region and provider, more may cost extra.Source | US West (California), US East (Virginia), EU West (Amsterdam) and Southeast Asia (Singapore). None in South America.Source |
| Run in your own cloud | Yes, with Connect: the same platform in your AWS, GCP, Azure or OCI account, or on-premise. 12-month minimum and a one-time setup fee.Source | Bring your own cloud is listed only on the Enterprise plan (custom pricing, contact sales).Source |
| Autoscaling | CPU- and memory-based autoscaling for app environments, enabled per account on request. Scale to zero is Connect-only.Source | Automatic vertical scaling up to plan limits. Horizontal scaling is a replica count you set; the docs describe no metric-based horizontal autoscaling.Source |
| Logs and metrics | Included: logs kept 20 days, Grafana metrics, alerts to Slack, Teams, PagerDuty, email or webhooks, and OpenTelemetry tracing (enabled per account).Source | Log explorer kept 7 days on Hobby, 30 on Pro, up to 90 on Enterprise. CPU, memory, disk and network metrics; tracing is in preview.Source |
| Preview environments | Branch previews per pushed branch with the Quave ONE CLI and GitHub Action, cleaned up after a TTL.Source | PR environments created automatically per pull request and deleted on merge or close.Source |
| Support | Email support, Monday to Friday, 9AM–5PM São Paulo time, usually answered within an hour. 24/7 on-call for infrastructure outages on every plan.Source | Community support on Free and Hobby. Pro gets direct help through the forum, usually within 72 hours. No email support.Source |
| Compliance | Working towards SOC 2 (preparing for the audit). Customers with SOC 2 and ISO 27001 certifications run on Quave ONE; with Connect, workloads stay in your own cloud account.Source | SOC 2 Type II and SOC 3. HIPAA BAA as a committed-spend add-on.Source |
| AI agents (MCP) | Remote MCP server included: AI clients can inspect logs, review deployments, manage environments and run approved Jobs, within your account permissions.Source | Official hosted MCP server (mcp.railway.com) and plugins for common coding agents.Source |
Every fact on this page links to the public page it came from: our own docs for Quave ONE, and the vendor's own pricing or docs pages for everyone else. We checked them on November 10, 2026. Prices and features change, so check the source before you decide. If something here is wrong or out of date, email support@quave.one and we will fix it.