# Steel

> Open-source browser API that spins up fleets of cloud browsers for agents to research, scrape, shop and fill forms

Source: https://dontaskagain.com/steel
Website: https://steel.dev/?utm_source=dontaskagain
Pricing: Freemium
Categories: [Sandboxes](https://dontaskagain.com/categories/sandboxes.md), [Browser agents](https://dontaskagain.com/categories/browser-agents.md), [Open source](https://dontaskagain.com/categories/open-source.md)
Listed: 2026-10-24
Updated: 2026-10-24

Steel is an open-source browser API for AI agents that controls fleets of real cloud browsers. Agents use it to navigate, click and type through websites to research, scrape data behind logins, compare prices, fill forms and run end-to-end tests.

**Key features:**

- Sessions API that starts a browser in under a second.
- Existing Playwright, Puppeteer or Selenium scripts connect without changes.
- Resumable sessions that keep cookies, tabs and state.
- Proxies and built-in CAPTCHA solving to reduce bot detection.
- Browser tools such as a /scrape endpoint for page content.
- Thousands of concurrent sessions for large jobs.

Steel offers Python and Node.js SDKs, a REST API and MCP access, and the core browser infrastructure is open source. The hosted Launch plan is free with $30 of one-time credits and 10 concurrent sessions. Scale is $250/month plus usage, with $100 of monthly credits, 100 concurrent sessions and a HIPAA-ready BAA. Enterprise adds a stealth browser, sessions up to 24 hours, SSO and SLAs.

## More sandboxes

- [Browserbase](https://dontaskagain.com/browserbase.md): Cloud browser platform that lets AI agents log in, navigate, fill forms and extract data from any website at scale
- [Hyperbrowser](https://dontaskagain.com/hyperbrowser.md): Cloud browsers, sandboxes and web data APIs for AI agents, plus managed agents that run browser tasks from a prompt
- [Anchor Browser](https://dontaskagain.com/anchor-browser.md): Cloud browser infrastructure that runs fleets of managed Chromium agents to automate authenticated web tasks at scale
- [Cua](https://dontaskagain.com/cua.md): Shared virtual desktops where you and computer-use agents like Claude Code or Codex work side by side
- [E2B](https://dontaskagain.com/e2b.md): Cloud sandboxes that give each AI agent an isolated microVM with a shell, filesystem, browser and network
- [Daytona](https://dontaskagain.com/daytona.md): Sandbox infrastructure for running AI-generated code in isolated, stateful environments that start in under 90 ms
- [Modal](https://dontaskagain.com/modal.md): Serverless AI cloud for inference, training, batch jobs and scalable sandboxes that give agents their own computers
- [Runloop](https://dontaskagain.com/runloop.md): Micro-VM sandboxes called Devboxes for building, evaluating and running AI coding agents at scale