Steel
An open-source browser API gives AI agents control of Chrome, with persistent sessions, page extraction, and local or self-hosted Docker deployment.
Steel is a browser API for developers building AI agents that interact with websites. It supplies real Chrome sessions for tasks involving clicks, forms, logins and pages that rely on JavaScript.
- Persistent sessions: Agents can keep cookies and local storage across requests to continue tasks without starting with a fresh browser each time.
- Automation compatibility: Connect through Puppeteer, Playwright or Selenium, with Python and Node.js SDKs available for application code.
- Page extraction: Retrieve HTML, Markdown or readable page text, and capture screenshots or PDFs for research, scraping and records of browser activity.
- Browser customization: Use proxies and custom Chrome extensions for workflows that need different network routes or additional browser functions.
- Session inspection: Request logs and a session viewing interface help developers investigate an agent's browser activity.
The open-source browser server runs locally or on your own server through Docker under the Apache 2.0 license. Selenium support doesn't include every feature of the other browser sessions.
Steel also offers a hosted service where browsers run in its cloud. That service includes automatic CAPTCHA solving, persistent agent profiles, readable traces and session recordings. Its live viewer lets a person take control for steps such as two-factor authentication or payments, then return control to the agent.
