Steel Browser
Steel Browser: 🔥 Open Source Browser API for AI Agents & Apps. Steel Browser is a batteries-included browser sandbox that lets you automate the web without worrying about infrastructure.
Last verified:
What is Steel Browser?
Steel Browser is Steel is an open-source headless browser API purpose-built for AI agents and applications that need to interact with the web. It lets developers control fleets of browser sessions in the cloud via API or Python/Node SDKs, eliminating the need to manage browser infrastructure, handle proxies, or implement CAPTCHA solving themselves. Steel provides sub-second session startup times and supports sessions running up to 24 hours with persistent cookies and local storage.
Key features include built-in auto CAPTCHA solving, rotating residential proxies with browser fingerprint randomization, stealth mode configurations to avoid bot detection, session persistence for maintaining authentication state, live session viewer for debugging, and optimized page formats that reduce LLM token usage by up to 80%. Steel integrates seamlessly with Puppeteer, Playwright, and Selenium with just a one-line code change to run in the cloud.
Steel is designed for AI engineers building AI agents, shopping assistants, web scraping tools, RPA software, sales automation tools, QA software, customer service agents, and foundational model training pipelines. It serves developers who need reliable browser automation without infrastructure headaches, from hobbyists prototyping projects to enterprises running thousands of concurrent browser sessions.
Steel Browser pricing
Pricing model: Freemium
Steel offers a free Hobby plan with $10 in credits per month (100 browser hours), no CAPTCHA solves, no proxy bandwidth, 15-minute max session time, 5 concurrent sessions, 500 daily requests, 1 req/sec limit, 24-hour data retention, and community support only. Starter plan is $29/month with $29 credits (290 browser hours, 2.9GB proxy bandwidth, 7.2k CAPTCHA solves), 1-hour max session, 10 concurrent sessions, 1,000 daily requests, 2 req/sec, 2-day retention, email support. Developer plan is $99/month with $99 credits (1,238 browser hours, 12GB bandwidth, 28k CAPTCHA solves), 6-hour max session, 20 concurrent sessions, unlimited requests, 5 req/sec, 7-day retention. Pro plan is $499/month with $499 credits (9,980 browser hours, 166GB bandwidth, 166k CAPTCHA solves), 24-hour max session, 100 concurrent sessions, unlimited requests, 10 req/sec, 14-day retention, plus dedicated Slack. Enterprise offers custom pricing with higher limits, discounted rates, dedicated clusters/APIs, and unlimited session concurrency. All paid plans include pay-as-you-go overages up to 3x monthly credits. Browser hours billed by minute rounded up at $0.10/hr (Hobby/Starter), $0.08/hr (Developer), $0.05/hr (Pro). CAPTCHA solves cost $4/1k (Starter), $3.5/1k (Developer), $3/1k (Pro). Proxy bandwidth costs $10/GB (Starter), $8/GB (Developer), $5/GB (Pro). Annual billing saves 20%.
Steel Browser pros
- Open-source browser API that can be self-hosted with Docker
- Sub-second session startup time (<1s) when client is in same region
- Sessions can run up to 24 hours with persistent state
- Built-in auto CAPTCHA solving for common CAPTCHA types
- Rotating residential proxies from global network included
- Browser fingerprint randomization to avoid bot detection
- Stealth mode configurations out of the box
- Save and reuse cookies/local storage across multiple runs
- Live session viewer for debugging live or recorded sessions
- Reduces LLM token usage by up to 80% with optimized formats
- One-line change to run existing Puppeteer/Playwright/Selenium code
- Python and Node.js SDKs available
- Extract page data as cleaned HTML, markdown, PDFs, or screenshots
- Auto sign-in for auth-walled websites and apps
- Fully managed cloud infrastructure with automatic scaling
- Pay-as-you-go overages up to 3x monthly credit on paid plans
- Unlimited team members per account on all plans
Steel Browser cons
- Free Hobby plan limited to 15 minutes max session time
- Free plan only allows 5 concurrent sessions
- Free plan has 500 daily request limit
- Free plan only 1 request per second rate limit
- Free plan data retention only 24 hours
- No CAPTCHA solving included on free Hobby plan
- No proxy bandwidth included on free Hobby plan
- Email support only available on paid plans (not free)
- Primarily designed for cloud usage, self-hosting requires Docker knowledge
- Not ideal for offline operation without internet dependency
Frequently asked questions about Steel Browser
What is Steel and what does it do?
Steel is an open-source headless browser API purpose-built for AI agents and applications. It lets AI engineers control fleets of browser sessions in the cloud via API or Python/Node SDKs, extract page data as cleaned HTML/markdown/PDFs/screenshots, access data behind logins with persistent cookies and automatic sign-in, render complex client-side content with JavaScript execution, bypass anti-bot measures with rotating proxies and CAPTCHA solving, and reduce token usage by up to 80% with optimized page formats. All infrastructure is fully managed and ready to scale.
Is Steel open source?
Yes, Steel is proudly open source. You can run sessions locally or self-hosted using their Docker container available on GitHub at github.com/steel-dev/steel-browser. The steel-browser repository is a batteries-included browser instance that includes all the functionalities of their hosted service, with Python/Node SDKs, session viewer, and options for one-click deployment. You can choose to run it locally or host it yourself.
How long can a browser session run on Steel?
Each session can run for up to 24 hours on Steel. However, the maximum session time depends on your plan: Hobby (free) plan allows 15 minutes, Starter plan allows 1 hour, Developer plan allows 6 hours, Pro plan allows 24 hours, and Enterprise has custom limits. Sessions maintain cookies, local storage, and authentication state across requests, enabling complex workflows like monitoring dashboards requiring login, multi-step form submissions, and long-running data extraction tasks.
Does Steel handle CAPTCHA solving?
Yes, Steel includes built-in auto CAPTCHA solving for common CAPTCHA types that keeps your automation flowing. The Hobby (free) plan does not include CAPTCHA solves, but Starter plan includes 7,250 CAPTCHA solves, Developer includes 28k solves, and Pro includes 166k solves. Additional CAPTCHA solves are billed at $4/1k (Starter), $3.5/1k (Developer), or $3/1k (Pro). For complex CAPTCHA types that require special handling, Steel provides CLI commands like 'steel browser captcha solve --session' for agents to handle them.
What automation frameworks does Steel support?
Steel works seamlessly with Puppeteer, Playwright, and Selenium. You can easily run your existing Puppeteer, Playwright, or Selenium code in the cloud with just a one-line change - replacing puppeteer.launch() with puppeteer.connect(). Steel provides official SDKs for Python and Node.js, and is always working on adding more framework integrations. The platform abstracts away browser infrastructure while maintaining compatibility with your favorite automation frameworks.
How fast is session startup on Steel?
Steel achieves sub-second session startup times - average session starts in less than 1 second when the client is in the same region. This is significantly faster than self-managed browsers which typically take 3-5 seconds for cold starts. The quick-start capability is one of Steel's key advantages, with benchmarks showing 500ms for Steel vs 5000ms for competition.
How does Steel help avoid bot detection?
Steel includes enterprise-grade anti-bot evasion techniques out of the box: residential proxy rotation from a global network, automatic CAPTCHA solving for common types, browser fingerprint randomization, human-like behavior simulation, and stealth mode configurations. The platform provides simple controls to never worry about getting flagged as a bot again. Proxy and browser fingerprinting management is included in all paid plans, with bandwidth limits of 2.9GB (Starter), 12GB (Developer), or 166GB (Pro) per month.
Can I debug my browser sessions on Steel?
Yes, Steel provides world-class observability with a Session Viewer that lets you view and debug live or recorded sessions. You can get a live view URL for active sessions and open it in your web browser to see what the agent is doing. The platform also provides session replays and embeddings for debugging. Session lifecycle commands like 'steel browser live' print the live view URL for the active session, and 'steel browser sessions' lists sessions as JSON for scripting.
Does Steel offer discounts for researchers or students?
Yes, Steel offers Research Grants in the form of free credits for those doing research for a school. You can apply by sending an email from your school email to [email protected] requesting grants. This is designed to support academic research involving browser automation and AI agents.
What are the rate limits and concurrency limits on Steel?
Rate limits and concurrency depend on your plan: Hobby (free) has 500 daily requests, 1 request/second, 5 concurrent sessions, and 15-minute max session. Starter has 1,000 daily requests, 2 req/sec, 10 concurrent sessions, 1-hour max session. Developer has unlimited daily requests, 5 req/sec, 20 concurrent sessions, 6-hour max session. Pro has unlimited requests, 10 req/sec, 100 concurrent sessions, 24-hour max session. Enterprise has custom limits. All paid plans include pay-as-you-go overages up to 3x monthly credit allocation.