
Public Browser
Local installChrome over CDP for AI agents. Blind benchmark vs agent-browser: -33% tokens, -33% cost, -32% time
332 servers in Browser Automation

Chrome over CDP for AI agents. Blind benchmark vs agent-browser: -33% tokens, -33% cost, -32% time
Let AI drive Swing and JavaFX apps — see the UI, click, fill forms, read tables. No API needed.

Use Reddit, Zhihu, X and WhatsApp through your signed-in browser with compact, on-demand MCP tools.
Screen observation, OCR and permission-controlled app input over MCP for Apple silicon Macs.

Web scraping with stealth HTTP, real browsers, and Cloudflare bypass. CSS selectors supported.
Browser automation using the user's Chromium session — JS execution, auth, and anti-bot support.
Bridge a live Chrome browser to your agent: 150 tools for capture, extraction, and observation.
Windows computer use in the background for Antigravity: UI Automation, Win32 messages, read back.
Windows computer use in the background: UI Automation and Win32 messages, every action read back.
Android computer use over plain ADB: screen, gestures, SMS, notifications, files. No root, no app.
Control existing signed-in Chrome and Edge tabs through a locally paired extension.
Stealthy browser automation, testing, and web-scraping via CDP Mode.
Control visible, persistent Hronaut browser workspaces through a local MCP connection.
Read and act on Safari — history, tabs and reading list, writes off by default
Let your coding agent use the Firefox, Zen, or LibreWolf browser you already have open.
Headless browser for AI agents with Chrome's fingerprint that passes Cloudflare challenges.
Windows screenshots, OCR, and redacted screen recording for AI agents. No desktop app needed.
Real browser for any agent: Chrome/Edge/Brave via CDP, Firefox via BiDi. Refs, forms, console, net.
Screenshots, PDFs, visual diffs and accessibility audits, returned inline so a model can see them
SHAFT browser automation, capture, and offline deterministic failure diagnosis
104 tools: drive YOUR real Chrome via CDP + browse/install/publish Threads in chat. webloom.run.
Open live web and mobile app sessions from an e2e project to inspect the app and try locators.
Let agents see and drive a Mac: screenshots, clicks, typing, apps and virtual displays.
Sealed X11/Wayland sessions to see, drive and check real GUI programs: a11y, pictures, sound.
Create, launch, check and drive Nullprint multi-profile browser profiles from any MCP agent.
Drive your real Chrome over MCP: real logins and cookies, multi-tab automation, deny-all by default.
Let your AI agent see, use and control the apps on your phone: screen as text, tap by label.
Let AI agents see, tap, type, scroll, screenshot and debug your Expo app on a real phone.
Real Chromium browser for AI agents: compact outlines, diff-only replies, real input, you approve.