Railway

Automatically deploys Docker containers from GitHub repositories on push or pull request events, with branch-based routing and automatic preview environment creation. Railway monitors GitHub webhooks, builds container images using Railpack (automatic configuration) or custom Dockerfiles, and routes traffic based on branch names. Preview environments are automatically torn down on merge, enabling zero-configuration staging workflows without manual environment management.

Charges for CPU and memory consumption at granular per-second intervals ($0.00000772 per vCPU/second, $0.00000386 per GB/second) rather than fixed instance sizes, with automatic vertical scaling on Pro/Enterprise tiers that adjusts CPU/RAM allocation based on real-time workload demand. Horizontal scaling supports up to 50 replicas with automatic load balancing, enabling cost-efficient burst handling for variable-load AI services without pre-provisioning peak capacity.

Exposes a GraphQL API with 100+ methods enabling programmatic deployment, configuration, and monitoring of Railway services. The API is the same interface powering the Railway console, enabling infrastructure-as-code workflows and custom automation. API authentication uses Railway tokens, and responses include deployment status, service metrics, and configuration details.

Railway CLI (25+ commands) enables deployment of local repositories without GitHub integration, supporting manual pushes and local testing workflows. CLI commands include service creation, configuration management, log streaming, and deployment status checks. Local deployments are useful for testing before pushing to GitHub or for CI/CD systems that don't integrate with GitHub.

Collects structured JSON logs from all services with configurable retention (7 days Hobby, 30 days Pro, 90 days Enterprise) and supports log forwarding to external systems. Logs are queryable and filterable by service, timestamp, and log level, enabling debugging and audit trails. Log forwarding enables integration with external log aggregation platforms (e.g., Datadog, Splunk) for long-term retention.

Provides 2,000+ pre-built deployment templates for common services (databases, frameworks, tools) that can be customized and deployed with one click. Templates are shareable and customizable, enabling teams to standardize service configurations and reduce deployment time. Templates include pre-configured environment variables, resource allocations, and health checks.

Provides a real-time visual project canvas showing all services, databases, and connections with drag-and-drop interface for managing infrastructure. Enables team collaboration with shared project access and real-time updates. Available only on Pro/Enterprise tiers. No explicit documentation on concurrent editor limits, conflict resolution, or audit trails.

Provisions managed database instances (PostgreSQL, MySQL, MongoDB, Redis) as Railway services with automatic backups, point-in-time recovery, and connection pooling. Databases are deployed as containers within the same Railway project, enabling zero-configuration networking between services via internal DNS (service-to-service communication over private 100 Gbps network). Persistent volumes up to 5 TB store database files with automatic IOPS provisioning (3,000 read/write operations per second standard).

Executes containerized cron jobs on a schedule with 5-minute minimum interval granularity, enabling periodic ML tasks like model retraining, batch inference, and data pipeline runs. Jobs are deployed as Railway services with the same container infrastructure as web services, sharing access to databases and environment variables. Execution logs are retained for 7 days (Hobby), 30 days (Pro), or 90 days (Enterprise), enabling debugging of failed training runs.

Deploys services across 4 geographic regions (US East, US West, Europe West, Southeast Asia) with automatic load balancing and failover. Enterprise tier supports concurrent multi-region deployments with a single configuration, routing traffic based on geographic proximity or custom rules. Internal networking (100 Gbps private) enables low-latency service-to-service communication within regions, while public egress (10 Gbps) is shared across replicas.

Manages service configuration through encrypted environment variables with support for variable references (e.g., $DATABASE_URL), shared variables across services, and automatic secret encryption at rest. Variables are scoped to individual services or shared project-wide, enabling centralized configuration management without duplicating credentials across multiple services. Secrets are encrypted using Railway's key management system and never exposed in logs or deployment artifacts.

Provides drag-and-drop customizable dashboards displaying real-time metrics (CPU, RAM, disk, network usage) for services and databases. Metrics are collected automatically from all Railway services without instrumentation, enabling quick visibility into infrastructure health. Dashboards support custom time ranges, metric aggregation, and export capabilities for post-incident analysis.

Sends configurable alerts to email, Slack, or Discord when service metrics exceed thresholds or usage limits are approached. Alerts support low/high urgency levels, enabling different notification channels for critical vs. informational events. Usage alerts can be configured with soft caps (warning) and hard caps (service termination), preventing unexpected billing surprises.

Configures custom HTTP health check endpoints with adjustable timeout and interval settings, automatically restarting services that fail health checks. Health checks are performed at configurable intervals (default not documented) and services are marked unhealthy after consecutive failures, triggering automatic restart. Failed health checks are logged and visible in Railway's monitoring dashboard.

Provides persistent storage volumes up to 5 TB per service with automatic IOPS provisioning (3,000 read/write operations per second standard tier). Volumes are mounted to service containers at specified paths, enabling stateful services like databases and file caches. Storage is billed separately at $0.00000006 per GB/second, and volumes persist across service restarts and deployments.