Before we dive into this week’s changelog (object storage edition), we’ve got something new we think you’ll love: 

We’re rolling out a brand new metal-based build environment (in beta) to make your builds faster than ever. It’s a big step forward in performance, and we’d love your help putting it through its paces.

Service Settings → Build → Enable Metal Builders

 to turn it on. Try it with your existing projects, compare build times, and share your feedback. Every insight helps us polish things up and get this ready for prime time. If you hit any issues, 

And that’s just the start. This week we also shipped improvements to the dashboard, 

But that’s not all, we’re releasing Stripe Connect to cash out bounty winnings, clearer error messages in the Data tab, and adding replica-level metrics to help you keep a closer eye on your services.

Now let's get to what you've all been waiting for, the main event, Object Storage!

We’re introducing a new infrastructure primitive that you can use when building your apps: object storage. No need to reach for 3rd-party services when wanting to store files.

 – Works with existing S3 tools and clients

 – Buckets are private by default. If you want public access, you’ll need to deploy a public proxy to serve those files.

 – Object storage integrates with environments. Every new environment you spin up will have its own bucket

 for users on the Hobby or Pro plans (currently not part of the free plan to prevent abuse). In terms of limits, workspaces can store up to 1 TB across all workspace buckets


To get started, you can right-click in your project canvas and choose the “Bucket option”

This is just the start, we plan on rolling out more features soon! If you run into any issues, drop us a note in this 

 are open requests from the community that anyone can jump in and solve. These can be requests for:

 - share reusable project templates to help others get started quickly

 - lend a hand by solving tricky problems and help others debug issues

 - tell us what you’ve shipped on Railway and how you built it

Until recently, bounty winnings were awarded as Railway Credits. But here’s the big news: the bounty board just leveled up. You can now turn your bounty winnings into 

. Yup, your side quests now come with a direct deposit (via Stripe Connect) button.

Your earnings now show up right alongside templates and referrals on your 

We’re working hard on giving you a world-class experience for databases on Railway. And this week we’ve got a big batch of updates designed to make it easier to see what’s going wrong, and even better, what you should do to fix it. (Shout-out to 

Completed → We’ll open your deploy logs

Crashed → We’ll open your deploy logs

No deployment → You’ll be prompted to create one

Smarter deployment states – No more guessing what’s happening

Clearer help with missing variables – Instead of dumping the whole list of required variables, we now show you exactly which ones are missing and prompt you to configure them

Not set? → We’ll recommend the correct port and prompt you to add it

Set incorrectly? → We’ll suggest the right port and prompt you to fix it

And don’t worry, no connection attempts will be made if things aren’t set up properly.

TCP Proxy validation – We’ll check if your TCP Proxy is set correctly on the right application port

TCP Proxy autofill – When you’re adding a TCP Proxy for a database service, we’ll automatically suggest the recommended port for you

More useful error messages – Any errors returned from the API (failed authentication, connection reset, etc.) are now surfaced directly in the UI, so you know exactly what went wrong

Polished design details – Unified look and feel for loading states and checks in the data tab: consistent banners, buttons, and more

Logs panel fix – The logs panel now opens to the right state when navigated to with a hash

If you have any feedback about this update, reach out in 

Metrics are now available for all your replicas!

When you’ve got a bunch of replicas running for a service, it wasn’t clear which ones were pulling the weight and which ones were just coasting along. Now, you can dig into metrics for each replica and:

Check performance details for every single one

Spot when one replica is carrying more load than the rest

Figure out if it’s time to scale up or rebalance

Give it a spin, and if something feels off, let us know over at 

We improved the dashboard’s command palette redeploy action. Previously, the redeploy action always showed the fixed text "Deploy latest commit", which caused confusion for users when services were deployed from an image. Now, the label dynamically adapts to the service type: "Deploy latest commit" for repo-based services, "Redeploy service image" for image-based services, "Redeploy latest deployment" when no source is set but a deployment exists, and hidden entirely when no deployments are present. This makes the action clearer and provides more accurate guidance for users

 command now supports by adding fetching a specific number of lines with filters. Here’s a sample command 

railway logs --lines 10 --filter "@level:error AND rate limit"

 . This should improve the debugging experience when using the CLI

 command now supports pulling build logs for a specific build. This is useful for debugging failed builds

You can now list deployments and their statuses by running 

 command is also updated to document these changes

We shipped several improvements to the Railway CLI 

If you ask your coding agent to "find the latest failing build, look at the errors, commit a fix" it can use the new CLI capabilities to complete the task

Coding agents can now take advantage of the CLI logging update

The MCP server doesn’t return JSON by default to avoid quickly hitting the MCP token limit. Returning JSON is instead exposed as an option in tools that can return JSON

We shipped a couple of improvements to the Railway MCP server:

 are now easier to find, and you can filter by category and sort by max payout, newest bounties or based on the most recent activity

There are now typing indicators so you can see who’s actively replying to a thread. This should make it easier when multiple people are contributing to the same discussion

We shipped several improvements to Central Station:

https://res.cloudinary.com/railway/image/upload/v1758927948/2025-09-26_hmfz5v.png

Object Storage, Bounty Cash Withdrawals, 

Changelog #0257