railgate

Self-hostable localhost tunnel. Like ngrok, except the relay runs on infrastructure you own.

# One-time setup — deploys a relay to Railway, saves config
npx railgate setup

# Expose any local port
npx railgate http 3000

How it works

railgate is two pieces:

• A CLI client that runs on your laptop and connects to a relay over a WebSocket control channel.
• A relay that lives on the public internet, accepts incoming HTTP/WebSocket traffic, and forwards it through the control channel to your laptop.

npx railgate setup is the one-time provisioning step:

1. Generates a shared token between your CLI and your relay.
2. Opens Railway in your browser. You authorize railgate (standard OAuth, no client secret).
3. railgate creates a Railway project for you, deploys the relay image, waits for it to come up, captures the public URL, and verifies it responds correctly.
4. Saves everything to ~/.config/railgate/config.json (chmod 600).

After that, npx railgate http <port> just works.

Configuration

Effective config resolves as flags > env vars > saved config.

| Source | Format | |---|---| | CLI flags | --relay <url>, --token <value>, --subdomain <name> | | Env vars | RAILGATE_RELAY_URL, RAILGATE_TOKEN | | Saved | ~/.config/railgate/config.json (or $XDG_CONFIG_HOME/railgate/config.json) |

Alternative setup modes

# Skip the browser entirely — paste a relay URL + token you already have
npx railgate setup --manual

# Use the legacy "open a browser, paste back the URL" deploy flow
npx railgate setup --browser

Self-hosting the relay outside Railway

The relay is a small Node service. Pull it directly:

docker run -d \
  -p 3000:3000 \
  -e RAILGATE_TOKEN=$(openssl rand -base64 32) \
  -e BASE_DOMAIN=relay.example.com \
  -e PROTOCOL=https \
  ghcr.io/codyde/railgate-relay:latest

Then railgate setup --manual and point it at your relay.

Repo

https://github.com/codyde/railgate

License

MIT