I've been wiring browser automation into agent workflows for months, and the pattern kept repeating: the scraping logic was never the hard part — standing up the infrastructure was. Steel Browser is an open-source browser API built for agents, but self-hosting a headless Chrome setup with proper session isolation is a weekend project nobody wants to redo from scratch.
So I packaged it as a one-click deploy template on Railway. The detail I care about most isn't the deploy button — it's that every single route sits behind API-key auth middleware. No open endpoints, no accidentally public browser proxies. When an instance boots for the first time, the key gets injected as an environment variable, and the gateway rejects any request without it before it even reaches the browser layer.
The honest tradeoff as an operator: headless browsers are memory-hungry. A single session can pull 200–400MB depending on the page, so on a hobby-tier plan you're fine for light automation, but parallel sessions will push you into the next tier fast. I set a hard concurrency cap in the template config specifically to keep costs predictable.
If you want to spin up your own instance: https://github.com/INAPP-Mobile/steel-browser
Top comments (0)