---
title: "Railway Agent Skill, CLI & Dashboard Updates, Docs Feedback"
date: 2026-01-09
number: 0272
url: https://railway.com/changelog/2026-01-09-railway-agent-skill
---

# Railway Agent Skill, CLI & Dashboard Updates, Docs Feedback

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 [what should we ship in 2026 thread](https://station.railway.com/feedback/feedback-what-should-we-ship-in-2026-9a878de8). The suggestions keep rolling in, and we're reading every single one. If you haven't shared yet, it's not too late.

This week: we're shipping a [Railway Agent Skill](https://github.com/railwayapp/railway-skills) so your coding agents can deploy to Railway without leaving your editor, [the CLI ](https://docs.railway.com/guides/cli)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](https://station.railway.com/feedback/feedback-railway-docs-a0fdb300).

Let's get into it! 🚄

## Railway Agent Skill & Claude Plugin

[Video: Untitled]

Making Railway coding-agent-friendly has been a priority for us. We've already shipped the [Railway MCP Server](https://github.com/railwayapp/railway-mcp-server), which works well — but now we're adding something new: [Agent Skills](https://agentskills.io/home).

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.

We've built the [Railway Skill](https://github.com/railwayapp/railway-skills) so your coding agent can:

- Spin up projects
- Manage services
- Pull metrics
- Manage environments and variables
- …and a lot more

If you’re using Claude Code, you can install the Railway plugin. Make sure you’re on the latest version (you can run `claude update`) and run the following commands:

```bash
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[ skill directory](https://github.com/railwayapp/railway-skills/tree/main/plugins/railway/skills) into your project, and you're set.

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](https://station.railway.com/feedback/feedback-railway-agent-skill-44a00d87).

## CLI Updates

[Image: The Railway CLI keeps getting better]

Under the hood, the Railway Skill uses the[ Railway CLI](https://docs.railway.com/guides/cli) 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:

**New commands:**

- `railway restart` — Restart deployments without triggering a full redeploy
- `railway delete` — Delete projects directly from the CLI
- `railway upgrade` — Update the CLI to the latest version (now with Bun support)

**New flags:**

- `--json` for `railway link` — Machine-readable output, perfect for CI pipelines
- `--json` for `railway volume list` — Get volume info in JSON format
- `--print` for `railway open` — Print the URL instead of opening it in a browser
- `--project` for `railway run` and `railway up` — Run commands against a specific project without linking
- `--since` and `--until` for `railway logs` — Filter logs by time range
- `--latest` for `railway logs` — Jump straight to the most recent deployment's logs
- `-set-from-stdin` for `railway variables` — Pipe variable values from stdin
- `y/--yes` for `railway unlink` — Skip confirmation prompts for scripting
- `local` alias for `railway run` — A shorthand that just makes sense

**Better error messages:**

- Clearer feedback when a deployment can't be redeployed
- Improved SSH error messages for serverless services
- More helpful `InvalidRailwayToken` error messages

We'd love to hear what's still missing. Drop your CLI feedback [in this Central Station thread](https://station.railway.com/feedback/feedback-railway-cli-0e5ab6cd).

## Dashboard Quality of Life Improvements

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.

### Editable Canvas Arrows

[Video: Tweak route arrows between services to get the perfect canvas]

[*New in Priority Boarding: *](https://railway.com/account/feature-flags)[Editable Canvas Arrows](https://railway.com/account/feature-flags)

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:

- **Double-click** an arrow to enter edit mode
- **Drag** the start or end points to reposition connections
- **Click** anywhere on the arrow to add anchor points, then drag them into place
- **Alt-click or double-click** an anchor point to remove it

Your canvas, your layout. If the feature flag isn't already enabled, head to your [feature flag settings](https://railway.com/account/feature-flags) and turn on **Editable Connection Arrows**.

### Improved Delete Resource Dialog

[Image: See exactly what you're about to delete]

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.

### Buckets

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.

### Friendlier Container Image Errors

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:

- **Missing tag** — The tag you specified doesn't exist on the image
- **Missing architecture** — The image doesn't have a `linux/amd64` variant (e.g., `arm64v8/nginx` won't work on Railway's infrastructure)
- **Invalid user** — The user specified in your Dockerfile doesn't exist; you'll need to create it before switching to it
- **Invalid registry auth** — 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.

[Let us know](https://station.railway.com/new?type=feedback) If there’s anything you’d like to see in particular for the dashboard.

## Help Us Improve the Railway Docs

[Image: Help us make the Railway docs better]

All of the above features are only useful if you can find out how to use them. The [Railway docs ](https://docs.railway.com/)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](https://station.railway.com/feedback/feedback-railway-docs-a0fdb300). Everything helps.

## Fixes and Improvements

- 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.