Directory · Last checked 25 Oct 2026

Infrastructure for computer-use agents

If you build an agent on a model API, it needs a safe place to click: a cloud browser or a sandboxed computer. Grouped by job, alphabetical within each group.

Listings are free and unpaid. Pricing and features change often, so check each vendor's site. Suggest a tool or correction.

Cloud browsers

VendorWhat it providesNotes
Anchor BrowserCloud browsers for AI agentsFocus on authenticated, security-sensitive work
AirtopCloud browsers controlled with natural languageAimed at agents that work on logged-in sites
Bright DataBrowser API with built-in proxies and unblockingLarge proxy network; also sells web data
BrowserbaseManaged headless browsers for agents, with session observabilityMakes the open-source Stagehand framework; $40M Series B (2025)
BrowserlessHosted headless Chrome with an APIPlaywright and Puppeteer support; self-hosting option
Browser Use CloudHosted browsers and agents from the Browser Use teamStealth browsers, proxies, pay as you go
HyperbrowserCloud browsers paired with hosted agent runtimes
KernelFast-starting sandboxed browsers and browser poolsPer-second billing; managed agent authentication
NotteBrowser infrastructure and agent toolsPublishes the Browser Arena speed benchmark
SteelOpen-source browser API for agentsSelf-hostable; publishes an OSWorld leaderboard

Desktop sandboxes and virtual computers

VendorWhat it providesNotes
CuaOpen-source framework and containers for desktop agentsWindows, macOS and Linux sandboxes
DaytonaSandboxes for running agent workloadsListed as offering Windows desktops
E2BOpen-source cloud sandboxes, including E2B Desktop with a full GUIScreen, mouse, keyboard and live stream for agents

Agent frameworks

Covered in detail in open-source computer-use agents: Browser Use, Stagehand, Skyvern, Cua, Agent S, UI-TARS and Playwright MCP.

Web data for agents

Not computer use as such, but often used alongside it when an agent only needs page content rather than clicks.

VendorWhat it provides
ApifyMarketplace of ready-made scrapers and automation "actors"
FirecrawlTurns websites into clean, model-ready text and data
ScrapingBeeScraping API that handles browsers and proxies

How to choose

Sources: Mastra, Jul 2026, Proxidize, Notte, Simular, SiliconANGLE.

New tools, monthly