steel-browser

Steel Browser is an open-source browser API for AI agents and applications, released by steel-dev under the Apache 2.0 license. The README describes it as "The open-source browser API for AI agents & apps," and presents it as a way to build live web agents and browser automation tools.

The API manages sessions, pages, and browser processes so browsing tasks can be performed programmatically. It uses Puppeteer and CDP for control over Chrome instances and supports connecting via Puppeteer, Playwright, or Selenium. Session management maintains browser state, cookies, and local storage across requests, and the /sessions endpoint lets users relaunch the browser with custom options or extensions, such as a custom proxy, and reset browser state.

Other listed features include built-in proxy chain management for IP rotation, loading of custom Chrome extensions, request logging with a UI for viewing and debugging sessions, stealth plugins and fingerprint management, and automatic cleanup and browser lifecycle management. Browser tools expose APIs to convert pages to markdown, readability, screenshots, or PDFs.

The /scrape, /screenshot, and /pdf endpoints extract data from a webpage using a running Steel server, described as ideal for simple, read-only, on-demand jobs. Steel can be run locally from a pre-built Docker image or deployed to cloud providers, and Node and Python SDKs are available. The project is in public beta.