Smart Diagnosis for Failed Deployments

Railway's new Smart Diagnosis feature automates the debugging process for failed deployments. When a deployment fails, users can trigger AI-powered diagnosis that reads through build logs, configuration files, and recent code changes to identify the root cause. The AI agent then suggests fixes, including one-click applicable config patches. Users can either apply fixes directly or open an interactive conversation to dig deeper and work through fixes collaboratively.

IPv6 Outbound Connectivity

Railway now supports IPv6 for outbound connections from services, enabling connections to external services that only expose IPv6 addresses. The feature is opt-in and must be enabled per service in service settings, ensuring it doesn't unexpectedly change networking behavior. Inbound public connections continue to use IPv4.

Unified Agent Skills System

The agent skills system received a major overhaul, consolidating 13 separate skills (status, projects, deploy, service, domains, environments, databases, templates, metrics, and more) into a single installable skill called use-railway. The unified approach improves usability for new users and makes agent evaluation more efficient by automatically matching user intent to the appropriate reference and loading only necessary components. Multi-step workflows can now compose into a single response without manual step invocation. Installation is available via:

curl -fsSL http://railway.com/skills.sh | bash

or

npx skills add railwayapp/railway-skills

Additional Improvements

Railway also shipped a new telemetry control command (railway telemetry <enable | disable | status>) for easier management of usage telemetry collection, and enhanced the project usage page with detailed cost breakdowns similar to the workspace view.