The new year's off to a strong start. This was our planning week, but we couldn't help ourselves and a lot still shipped. Thank you to everyone who dropped ideas in our 

. The suggestions keep rolling in, and we're reading every single one. If you haven't shared yet, it's not too late.

 so your coding agents can deploy to Railway without leaving your editor, 

picks up over a dozen new features and flags, and the dashboard gets a handful of quality-of-life improvements. We're also putting out a call to 

help us make the Railway docs truly world-class

Making Railway coding-agent-friendly has been a priority for us. We've already shipped the 

, which works well — but now we're adding something new: 

Agent Skills are a simple, open format for giving agents new capabilities. They're folders of instructions, scripts, and resources that agents can discover and use to work more accurately and efficiently. Unlike MCP, there's no server to run — local or remote — and they consume less of the LLM's context window. Think of them as the next evolution in agent tooling.

If you’re using Claude Code, you can install the Railway plugin. Make sure you’re on the latest version (you can run 

claude plugin marketplace add railwayapp/railway-claude-plugin
claude plugin install railway@railway-claude-plugin

The good news is that the Railway skill is not exclusive to Claude Code. Every major coding agent (e.g. Cursor, Codex, OpenCode, etc.) supports the Agent Skills spec. So all you need to do is  clone or copy the

The Railway Agent Skill is still early — expect some rough edges. We're actively improving the Skill and want to hear what's working and what's not.

Let us know in this Central Station thread

Under the hood, the Railway Skill uses the

 and API to do its work. So naturally, we've been leveling up the CLI to match. The philosophy is simple: anything you can do in the dashboard should be doable from the terminal. Here's what shipped over the past couple of releases this week:

 — Restart deployments without triggering a full redeploy

 — Delete projects directly from the CLI

 — Update the CLI to the latest version (now with Bun support)

 — Machine-readable output, perfect for CI pipelines

 — Print the URL instead of opening it in a browser

 — Run commands against a specific project without linking

 — Jump straight to the most recent deployment's logs

 — Skip confirmation prompts for scripting

Clearer feedback when a deployment can't be redeployed

Improved SSH error messages for serverless services

We'd love to hear what's still missing. Drop your CLI feedback 

While agent tooling and CLI work grab the spotlight, the dashboard hasn't been neglected. We've locked in the big projects for this quarter (more on that soon), but in the meantime, here's a batch of polish and fixes that make your day-to-day a little smoother.

Tweak route arrows between services to get the perfect canvas

You know the ritual: drag a service, nudge another, try to get the connection arrows looking clean. Sometimes it works, sometimes it’s just not perfect.

This changes today and you now you can take control. Connection arrows are fully editable:

 the start or end points to reposition connections

 anywhere on the arrow to add anchor points, then drag them into place

Your canvas, your layout. If the feature flag isn't already enabled, head to your 

When deleting multiple resources at once, you now see all of them listed in the confirmation dialog. Previously, bulk deletes required a bit of faith. Now you can scan the list and make sure you're not about to nuke something important.

A couple of small but useful improvements:

Bucket sizes now display in KB and bytes — not just MB. Previously, a nearly-empty bucket would show "0 MB," leaving you guessing. Now you get the actual size.

Environment syncing now includes buckets. When you sync environments, buckets will be created or deleted along with everything else. Previously they were left out, which meant manual cleanup. Not anymore.

Deploying container images can fail for a lot of reasons, and the errors aren't always helpful. We now surface user-friendly messages for the most common issues:

 — The tag you specified doesn't exist on the image

 — The user specified in your Dockerfile doesn't exist; you'll need to create it before switching to it

 — Your registry credentials aren't working

These show up instantly when a container fails to start or an image fails to pull. No more digging through logs for cryptic errors.

 If there’s anything you’d like to see in particular for the dashboard.

All of the above features are only useful if you can find out how to use them. The 

are due for an overhaul, and we want your input before we start.

What's been hard to find? What guides didn't click? What feature did you have to figure out on your own because the docs weren't there?

Whether it's a missing page, a confusing explanation, or just a small nit — we want to hear it. 

Drop your feedback in this Central Station thread

We fixed an issue where there was UI jank and incorrect patch application when discarding multiple staged changes in rapid succession. The experience is now smooth and reliable.

https://res.cloudinary.com/railway/image/upload/v1768008409/2026-01-09_anbvgx.png

Railway Agent Skill, CLI & Dashboard Updates, Docs Feedback

Changelog #0272