Steel Browser

Use it when

• You want self-hosted browser infrastructure for AI web agents instead of depending on a paid cloud browser vendor
• You need managed headless Chrome sessions with persistent cookies, proxy rotation, stealth fingerprinting, and a debug UI that existing Puppeteer, Playwright, or Selenium code can connect to

What it solves

Not the fit when

• evading bot detection to abuse, overload, or attack sites you are not authorized to automate
• high-level agent planning; pair with an agent framework or your own loop
• Selenium sessions do not support all CDP-based session features
• internet-facing production use of the default local Docker deployment without adding appropriate authentication, network isolation, and rate limits; the README labels Steel public beta
• writing the agent reasoning loop itself
• guaranteed captcha solving
• turning scraped html into structured data

Install

docker run -p 3000:3000 -p 9223:9223 ghcr.io/steel-dev/steel-browser

Invoke

POST http://localhost:3000/v1/sessions to create a browser session (proxy, extensions, cookies), then connect Puppeteer/Playwright/Selenium; or POST /v1/scrape, /v1/screenshot, /v1/pdf for quick one-shot actions

Alternatives

No reviewed alternatives recorded yet.