Deploy Skyvern Browser Agent
Skyvern AI browser agent with web UI, API, PostgreSQL and a password gate
Agent
Skyvern API
Just deployed
/data
Web
Skyvern UI
Just deployed
Skyvern
Just deployed
Data
PostgreSQL
Just deployed
/var/lib/postgresql/data
Deploy and Host Skyvern Browser Agent on Railway
Skyvern is an open-source AI agent that drives a real Chromium browser. You describe a goal in plain language, such as filling in a form, logging in to download an invoice, or extracting a table, and Skyvern reads the page, asks a language model what to do, and clicks, types and extracts until the goal is met. It comes with a web UI, a REST API, multi-step workflows, schedules and an encrypted credential vault.
About Hosting Skyvern Browser Agent
The template runs Skyvern v1.0.58 as four services from official images pinned by digest: the Skyvern API with its built-in Chromium browser and live browser stream, the Skyvern web UI, PostgreSQL 17, and a small Caddy gateway that is the only public service. The API keeps screenshots, recordings, downloads, browser sessions and the credential vault on a 5 GB volume; PostgreSQL has its own 5 GB volume. The gateway asks for a generated password before any page, API call or screenshot loads, because Skyvern's self-hosted UI has no sign-in of its own. Task and navigation URLs that point at private addresses inside the project are refused, workflow code blocks are off, and telemetry is off. You bring a model provider key: OpenAI by default, with Anthropic, Gemini, Azure, OpenRouter and any OpenAI-compatible server available by changing variables.
Why Deploy Skyvern Browser Agent on Railway
Railway provides the HTTPS domain, private networking between the services, persistent volumes for runs and credentials, generated secrets, and health-checked deploys. The UI and API are wired together automatically, the browser and its live view work without extra setup, and the API is never exposed to the internet without the gateway password.
Common Use Cases
- Automate repetitive browser work on sites without an API: form filling, portal logins, invoice and report downloads
- Extract structured data from websites in plain language and fetch it through the REST API or SDKs
- Build scheduled multi-step workflows that combine browsing, extraction, file downloads and webhooks
Dependencies for Skyvern Browser Agent Hosting
The deployment uses Skyvern (skyvern.com), PostgreSQL, and Caddy.
Deployment Dependencies
- Skyvern API and UI v1.0.58 (Chromium, Xvfb and Playwright are included in the API image)
- PostgreSQL 17, and a 5 GB Railway volume each for PostgreSQL and the API
- An API key for a model provider (OpenAI by default); every task step calls the model
Sign in to the gateway as skyvern with the SKYVERN_PASSWORD value from the Skyvern service's variables. The UI opens already signed in to your organization.
Template Content
Skyvern API
public.ecr.aws/skyvern/skyvern:v1.0.58OPENAI_API_KEY
Your OpenAI API key (required). Every task calls the model.
Skyvern
caddy:2.11.9-alpinePostgreSQL
postgres:17.12-alpine