Hey there! We've found the following might help you get unblocked faster:

• 🧵 Second Railway-Provided Domain?

• 🧵 N8N MCP Server

• 🧵 Issue Hosting MCP Server on Railway

If you find the answer from one of these, please let us know by solving the thread!

I will keep adding more here as I think of them

• CRUD service level and project level environment variables

• Tail & search logs

• Flag: Read only mode

• Execute troubleshooting/debug commands directly on containers via railway ssh

I might be in the minority on this, but this seems like a waste of time. I would never trust an AI anywhere near production infrastructure for a variety of reasons. I'd much rather see Railway make progress on other areas of the platform, instead of feeding into the hype cycle.

Deploying services and adjusting those would be great. The MCP server doesn't have to be able to adjust every single thing in existence (please don't just map the API 1:1 to MCP with 50+ tools), but creating a service connected to a repo, debugging a deployment by accessing the logs and adjusting the build/start commands and a few other settings would be great. Creating a few of the standard database templates, connecting variables, etc. would be phenomenal. Serverless functions too.

A fairly smart log search (that doesn't instantly fetch hundreds lines of code) would be great for debugging.

I'd love to have some non-technical people in the team be able to create and deploy small projects of their own, and if they don't have to dive into railway to worry about the deployment, that would be phenomenal. For the more technically inclined people, debugging by having the AI find it's way through logs would be great.

Hey @reece! A few quick thoughts:

would never trust an AI anywhere near production infrastructure for a variety of reasons

Totally valid concern. That’s exactly why the MCP server won’t support any destructive actions (e.g. it won’t be able to delete services, functions or databases). These things can be done from the CLI, API, or Dashboard.

I'd much rather see Railway make progress on other areas of the platform

Progress is absolutely happening across the board, both on the product side and in our underlying infrastructure. This particular project is something only I’m building, and I figured it was worth getting early feedback.

instead of feeding into the hype cycle.

In my opinion things are being somewhat overhyped for sure. That said, we’re seeing more users trying to integrate Railway into their workflows with tools like Cursor, VSCode, and Windsurf. There’s even a community-maintained MCP server already out there, and folks are actively trying to prompt AI to automate the Railway CLI. So this isn’t us trying to jump on a hype train, but rather about supporting an emerging use case with a better experience.

Really appreciate you taking the time to share your thoughts.

CRUD service level and project level environment variables

Do you mind sharing more details? What type of updates are you thinking of

Flag: Read only mode

Can you expand on this a bit more? Read only access for what exactly?

Execute troubleshooting/debug commands directly on containers via railway ssh

This one IMO is way too risky given that AI can hallucinate and run destructive commands. (e.g. "hmm, this doesn't seem to work, I"ll delete the entire database and start over")

That's great feedback. Thank you for sharing

> please don't just map the API 1:1 to MCP with 50+ tools
It won't be. The tools we'll include will mainly focus on: going from 0 -> live deployment, iterate by being able to pull info about a project and debugging (e.g. pull logs)