Good morning, afternoon, and evening, and tomorrow!
We aim to please here at the Railway corporation and we hope that this Changelog will make you happier than baseline.
A macro note, we hear you on the annoyance about builds. We've been scaling that Infra up with demand, a lot of the new announcements we have here use pure VMs for your workloads. The existence of a feature doesn't preclude the existence of maintenance and improvement of the system. That said, we are speeding towards a world where builds don't exist and your application is just live (more on that soon).
On a separate note, Engineer Victor has been hard at work improving the Infra as Code experience on the platform. So much so, we outright called it "Infra as Code," not to be confused with our older "Config as Code" that we shipped in 2022.
We know.
Anyway, Config as Code is going away in favor of our new IaC SDK. Starting August 28, services in new projects will no longer be deployed using Config as Code (new services in existing projects should be fine), with the old system going away on December 1.
You can read more about the migration in the docs.
With that out of the way...
Let's talk Changelog #0304
Railway for Everyone
Deploy without an account"Wow, author, what an interesting feature announcement title!"
You are correct reader.
Deployment is still the slowest step. For those who are VPS native, your agent still has to spend 100s of turns getting your server on. For those used to platforms like Vercel already, you have to wait for things like GitHub and CI. (Or builds...)
So we've been working on https://dev.new - a prototype app builder on Railway.
It uses our new VMs that power Cloud Agents... and this, where your VM can run 24/7 with your favorite harnesses.
After launching it, we are super pleased with the uptake so far... but what if we made it MORE accessible.
Anyway:
When you go on railway.com/new or dev.new...
You can now make sites, as well as DBs, without an account.
The catch?
You have 60 minutes to build and then claim. Internally, we call this anonymous provisions, and we plan to extend this to more and more to more parts of the product while we tune the limits to get the next 100 million builders on the platform.
Cloud Agents Everywhere
Railway Cloud AgentsRailway runs your workloads, and now... your agents.
At Railway, we think your laptop cycles should be preserved and you should offload your intensive agents loops somewhere else so you can multi-track drift.
We have a big update landed...
Now, on the Railway dashboard, you can kick off tasks in your coding harness of choice from the dashboard that then lands you right into a running VM that you can access anywhere.
...like your terminal.
Also thanks to Engineer Cody, we updated the CLI with this experience. Your CLI now has access to railway ca where you can see all of your running agents in real time and then you can switch between your sessions.
But thats not all.
You can then kick off and then pick up the sessions from your phone which is available on iOS now. (Which... implements libghosty) Web/Terminal/Phone... agents everywhere.
With the railway-mcp included in all sessions, your agents can then either open a PR or kick off new services and deployments right within your session so you can continue to evolve your project.
Edge rules
Edge rulesEdge Rules let you customize how Railway’s edge handles requests as soon as they reach our global network, before they reach your service.
Navigate to Service settings → Edge → Edge Rules to build conditions against client IP addresses and CIDR ranges, hostnames, paths, or request headers. Combine conditions using all, any, or none logic, nest condition groups, and choose how Railway handles matching requests:
-
Allow requests and skip the remaining rules
-
Block requests with a custom 4xx status code and response message
-
Challenge suspicious traffic with a verification page
-
Redirect to another hostname or URL using a 301, 302, 307, or 308 response, with optional path and query preservation
-
Override caching with a custom TTL or bypass the cache
Rules run from top to bottom on every request. Drag to reorder them, and toggle individual rules off without deleting them. The first matching allow, block, challenge, or redirect action determines what happens to the request. Cache overrides continue to later rules, so the same ruleset can combine caching behavior with request handling.
The editor validates the full ruleset before saving, highlights invalid conditions, and includes a JSON mode for advanced configurations. Once saved, changes propagate to Railway’s edge nodes worldwide within seconds.
Check out the Edge Rules documentation, give them a try, and share your feedback on Central Station.
High availability Redis
Redis HARunning a single Redis instance means a node failure can take the database offline. With Redis HA, Railway converts your existing service into a cluster backed by Redis Sentinel and HAProxy. Sentinel handles primary election and automatic failover, while HAProxy routes reconnecting clients to the current primary within seconds.
To get started, navigate to Database → Config → High Availability on your Redis service. Choose the number of Redis replicas and reverse proxies, then click Convert to HA.
Railway creates a backup, provisions the cluster as staged changes, and opens the cluster view for review. From there, you can monitor node health, identify the current primary, scale the cluster, perform a coordinated switchover, or revert to a standalone Redis service.
The conversion drops active connections and changes your connection endpoints. Railway updates variable references within your project automatically, but you'll need to update any hardcoded Redis connection strings.
MySQL HA is also available in beta through Priority Boarding behind its feature flag. It converts an existing MySQL service into a Group Replication cluster with automatic failover and HAProxy connection routing.
Read the Redis HA documentation and MySQL HA documentation, and share your feedback on Central Station.