Skip to content

Steel Browser

Open Source

Steel is an open-source browser API that lets teams control fleets of cloud browsers, with a managed cloud tier that adds stealth, proxies, and CAPTCHA solving for AI agents.

HIPAA
90

Overview

Steel is browser infrastructure for AI agents and automation teams that need real browsers at scale, available as an open-source server you can host yourself or as a managed cloud service. It targets developers building scraping pipelines, agentic workflows, and any task that requires a full browser rather than a headless HTTP client.

A Real Browser Behind One API Call

Steel's Sessions API starts an isolated Chromium instance in the cloud in under a second, with no cold-start penalty. Each session comes with its own proxy, cookies, and storage already configured, so agents do not have to assemble a browser environment before they begin work. Sessions can be resumed with the same cookies, tabs, and state, which matters for workflows that span multiple steps or run over long periods.

Compatibility With the Tools Teams Already Use

Steel works with Playwright, Puppeteer, and Selenium, and it ships Python and Node.js SDKs alongside a REST API. Because connections use a standard CDP WebSocket URL, any CDP-compatible tool can point at Steel without custom integration work. A one-line change to an existing Puppeteer script is often enough to route it through Steel's cloud browsers. Coding agents such as Claude, Codex, Cursor, and Gemini can also set up Steel and browse through it, and an MCP integration is available.

Stealth, Proxies, and CAPTCHA Handling

Bot detection is a persistent obstacle for browser automation, and Steel addresses it with a stealth browser that rotates fingerprints and simulates human mouse movement. CAPTCHA solving is available on Steel Cloud, covering Cloudflare Turnstile, PerimeterX press-and-hold, and image CAPTCHAs, with a CAPTCHAs API that reports progress on each challenge. Stealth and proxies prevent many CAPTCHAs from appearing in the first place. Proxy options include residential and datacenter connections, country-to-city targeting, dedicated IPs, and rotating or sticky sessions.

Observability and Human-in-the-Loop Control

Steel provides readable traces that present each run as plain steps, viewable in the UI or read back through the API. A live session viewer records every session at 1080p, and humans can take over a running session for 2FA, payments, or other uncertain steps before handing control back to the agent. Session replays let teams review completed runs. Persistent browser profiles keep cookies and session tokens, so agents can pick up signed-in workflows without logging in again, and each agent can maintain its own account across thousands of runs.

Self-Hosting and Enterprise Deployment

Steel Browser is open source on GitHub, so teams can run it themselves, inspect the code, and keep data inside their own network. Steel Outpost and self-hosting options support bring-your-own-keys and region selection. The managed Enterprise tier adds more than 1,000 concurrent sessions, higher rate limits, volume discounts, dedicated capacity, premium support, 2FA, SSO, HIPAA-signed BAAs, isolated sessions, and encryption in transit.

Scale and Operational Fit

Steel reports serving more than 1.5 million browser hours, 35 million sessions, and 4 trillion tokens scraped. The platform is designed for fleets of browsers rather than single-session scripts, and it includes a CLI for starting, steering, and stopping cloud browsers from a shell, with screenshots and PDFs captured as evidence of each run. Skills for browsing and debugging extend the same commands to agents that run shell commands.

Where Steel Fits

Steel is a strong fit for teams that want to avoid building and maintaining browser infrastructure themselves, particularly when stealth, proxies, and CAPTCHA handling are part of the requirement. The open-source server appeals to organizations with data-residency or customization needs, while the managed cloud tier suits teams that would rather pay for scale than operate it. The main trade-offs are plan-level limits: stealth browser and dedicated IPs are not included on the entry-level Launch plan, session time caps range from 15 minutes to 24 hours depending on tier, and proxy bandwidth and CAPTCHA solves are metered separately.

Core features

  • Sessions API that starts isolated cloud Chromium sessions in under a second
  • Compatibility with Playwright, Puppeteer, Selenium, and Steel's Python and Node.js SDKs over a standard CDP WebSocket URL
  • Stealth browser with fingerprint rotation and human-like mouse movement
  • CAPTCHA solving for Cloudflare Turnstile, PerimeterX press-and-hold, and image CAPTCHAs
  • Persistent browser profiles that retain cookies and session tokens across runs
  • Live session viewer with 1080p recording and human-in-the-loop takeover
  • Residential and datacenter proxies with country-to-city targeting and rotating or sticky IPs
  • Self-hosting options including Steel Outpost and bring-your-own-keys deployment

Use cases

  1. 1Running large-scale web scraping and data extraction jobs
  2. 2Powering AI agents that need to browse, sign in, and complete multi-step web workflows
  3. 3Automating form fills, payments, and 2FA flows with human takeover when needed
  4. 4Testing and debugging browser automation scripts across Playwright, Puppeteer, or Selenium
  5. 5Operating agent fleets that require isolated sessions and consistent profiles at scale

Strengths and things to check

Strengths

  • Open-source core with a managed cloud option, so teams can choose self-hosting or scale without infrastructure work
  • Broad compatibility with existing CDP-based tools and popular automation frameworks
  • Built-in stealth, proxy, and CAPTCHA handling reduce common bot-detection blockers
  • Observability features such as readable traces, session replays, and a live viewer simplify debugging
  • Enterprise options include SSO, HIPAA-ready BAA, dedicated capacity, and higher concurrency

Things to check

  • Stealth browser and dedicated IPs are not included on the entry-level Launch plan
  • Session time limits vary by plan, from 15 minutes on Launch to up to 24 hours on Enterprise
  • CAPTCHA solving and proxy bandwidth are metered separately, which can affect cost at scale
  • Self-hosting requires teams to manage their own infrastructure and updates

Pricing

Steel Browser is free and open source. Steel Cloud offers Launch with $30 in one-time free credits, no monthly fee, and usage-based pricing; Scale is $250 per month plus usage; Enterprise pricing is custom. Metered rates include browser hours at $0.10 (Launch) and $0.08 (Scale), proxy bandwidth at $10/GB (Launch) and $6/GB (Scale), CAPTCHA solves at $3 per 1,000 (Launch) and $1 per 1,000 (Scale), and browser tools (/scrape) at $5 per 1,000 on both plans. Plan limits and feature availability vary.

Basic information

Website
steel.dev
Added
Oct 10, 2026
Platforms
Web
Chinese & English
Not marked
API
Marked as supported
Self-hosting
Marked as supported
Best for
Engineering teams and AI agent developers who need scalable, real-browser automation with self-hosting or managed cloud options.
Information updated
Oct 16, 2026
Website status
Checking
Corrections · Privacy · Data requests?
Contact us

Features, pricing and availability may change. Always verify details on the official website. Maintained by steven · Last updated Oct 16, 2026 →

Related tools

All categories