
Steel provides browser infrastructure for developers whose AI agents need to navigate websites, fill forms, and collect data. Its API creates cloud browser sessions that existing automation libraries can control. Steel Browser is the open-source server; Steel Cloud supplies a managed service with additional capacity and operational features.
Open-source deployment and managed cloud operation serve different needs. Teams can inspect or host Steel Browser themselves, while the cloud plans meter browser time, proxy bandwidth, and other services. Self-hosting the server does not imply that every managed-cloud feature is included or that infrastructure costs disappear.
Create a session with the API, connect a browser controller, and run the navigation your application requires. Reusing session context lets a later run pick up cookies and state instead of starting from a fresh login. Python and Node.js SDKs support integration into existing agents.
Use the live viewer and recordings to investigate failed clicks or unexpected pages. Proxy and CAPTCHA options support difficult browsing workflows, but session duration and concurrency still depend on the chosen plan. The dedicated Stealth Browser feature is listed on Enterprise.
Steel presents use cases including web scraping, research, and browser-based testing. Assess it with sites your application actually needs: check login persistence, error recovery, recording quality, and costs across repeated runs. An isolated successful session is insufficient evidence of production reliability.
Launch browser time is $0.10/hour and proxy bandwidth $10/GB; Scale rates are $0.08/hour and $6/GB. CAPTCHA solves are metered separately. The one-time Launch credit expires after 90 days.
Steel at steel.dev provides cloud browser sessions for AI agents. Steel Browser is its open-source server; Steel Cloud is the managed service.
+2 more
+2 more