AI browser automation tools
Agents and libraries that drive a real browser: opening pages, clicking, typing and filling in forms from plain-language instructions. Of the 534 tools on our automating workflows page, 43 fit this narrower job. By pricing: 28 open source, 1 source-available, 9 free or freemium, 3 paid, usage-based or enterprise, 2 pricing on the maker's site. The most-starred open repos here are Crawl4AI (86,582 stars), Playwright MCP (38,849 stars), Lightpanda (37,901 stars). Most come from the browser/computer use (21), MCP servers (8), app builders (2) categories.
Updated
Crawl4AI
Crawl4AI is an Apache-2.0 web crawler that turns pages into clean, LLM-ready Markdown, extracts structured JSON and handles browser automation for RAG pipelines and agents, runnable yourself with Python or Docker. Crawl4AI Cloud offers the same capabilities as a hosted scrape, extract and search API with managed browsers and proxies, plus setup for Claude Code, Cursor and Codex.
Added
Playwright MCP
Playwright MCP is Microsoft's Apache-2.0 MCP server that gives LLMs browser automation through Playwright, using structured accessibility snapshots instead of screenshots so no vision model is needed. It plugs into Claude, Cursor, VS Code, Codex and other MCP clients via npx @playwright/mcp; for high-throughput coding agents Microsoft now also offers the token-leaner Playwright CLI with skills.
Added
Lightpanda
Lightpanda is an AGPL-3.0 headless browser written from scratch in Zig for AI agents, crawlers and test automation, not a Chromium fork. It executes JavaScript and exposes the Chrome DevTools Protocol, so existing Playwright and Puppeteer scripts can connect to it, and its own benchmark of 100 real pages shows about 16x less peak memory and 9x faster runs than headless Chrome. Install it via Homebrew or Docker, or use the hosted Lightpanda cloud.
Added
Skyvern
Skyvern is an AGPL-3.0 browser-automation platform that uses LLMs and computer vision to complete workflows on websites it has never seen, instead of brittle XPath scripts that break when layouts change. It offers a Playwright-compatible SDK with AI actions, a no-code workflow builder for non-technical users, and a hosted cloud version with proxies, CAPTCHA solving and 2FA handling.
Added
ego lite
Free macOS browser where Claude Code or Codex run automation tasks in their own isolated Spaces next to your tabs, reusing your logged-in sessions through an agent skill.
Added
Browser Use Web UI
Gradio web interface for the Browser Use agent that runs AI browser automation tasks with many LLM providers and your own browser.
Added
Bolt
Prompt-to-app builder that runs full-stack apps in the browser.
Added
Nanobrowser
Open-source Chrome extension agent that automates web tasks in your own browser.
Added
SeleniumBase MCP
SeleniumBase MCP is an MCP server that gives agents a way to handle stealthy browser automation, testing, and web-scraping via CDP Mode.
Added
Browser MCP
Browser MCP lets Claude, Cursor, Codex and other AI apps use your real browser — with your logins — in tabs you can watch and stop.
Added
Playwright
A Model Context Protocol server that provides browser automation capabilities using Playwright.
Added
Playwright Skill
Playwright automation skill for coding agents.
Added
Sheprd
Sheprd is an Apache-2.0 native workspace for terminals and AI coding agents on macOS, Linux and Windows, built in Rust with GPUI. It organises repositories, Git worktrees, tabs and splits, lets you chat with agents, review their changes and track pull requests and usage across local, SSH and cloud machines, and adds browser previews, automations and phone remote control on top of a Herdr terminal daemon.
Added
Skales
Personal AI agent for desktop and mobile that codes, automates the desktop and browser and runs scheduled tasks with local or cloud models.
Added
Qualitylayer
QualityLayer is a workflow layer for coding agents that keeps you in the loop: your agent (Claude Code, Codex, Cursor and others, on your existing login) drafts a plan with diagrams and clickable screens for you to approve, then writes tests before code, splits independent parts across agents and verifies each small change. You work in its desktop app, the browser or your own terminal, IDE or agent app.
Added
Ui.Vision RPA
Browser and desktop automation extension with OCR and image recognition, plus an MCP server so AI agents can create and run macros.
Added
Notte
Browser infrastructure for AI agents: stealth cloud browsers with CAPTCHA solving and proxies, a Playwright-based framework and serverless automations.
Added
Terminator
Desktop automation library and MCP server that lets AI assistants control any Windows app through the accessibility tree, like Playwright for the desktop.
Added
AIPex
AIPex: AI browser automation assistant, no migration and privacy first. Alternative to Manus Browser Operator、 Claude Chrome and Agent Browser
Added
Agent-E
Emergence AI's web agent that automates tasks in the browser through hierarchical planning.
Added
ClawBench
ClawBench is an open benchmark for browser AI agents on everyday tasks, measuring how well agents browse and complete real workflows.
Added
Browserable
Browserable is an MIT-licensed, self-hostable browser automation library for building AI agents that navigate sites, fill in forms, click buttons and extract information, reporting 90.4% on the WebVoyager benchmark. You start it with npx browserable or Docker Compose, add a key for Gemini, OpenAI or Claude and a remote browser provider such as Hyperbrowser or Steel, then run tasks from its dashboard, task API or JavaScript SDK.
Added
Opentabs
Opentabs is an MCP server that gives agents a way to handle browser automation clicks buttons.
Added
Libretto
An open-source CLI, debugging agents and browser tools for building and maintaining reliable web integrations, helping coding agents inspect real pages and write automations that keep working.
Added
MOVO
MOVO Community Edition is a self-hosted enterprise agent platform that takes DeepSeek Harness agents from experiments into production. It wraps the DSH runtime, skills, tools and MCP ecosystem in a deployable workspace with enterprise knowledge bases, deep research, content generation, vibe coding and browser automation, plus identity and access control, admin consoles, governance and file delivery. It starts locally with Docker in about five minutes.
Added
Interceptor
Browser automation and computer use for AI agents. One CLI and MCP server for your signed-in browser, native macOS apps, and real iPhones
Added
Bux
Bux is an open-source agent project that helps builders handle browser Use Box: a 24/7 Claude Code agent for Playwright-style browser automation.
Added
Shaft Mcp
Shaft Mcp is an MCP server that gives agents a way to handle SHAFT browser automation, capture, and offline deterministic failure diagnosis.
Added
Open Browser Use
Open Browser Use is an open-source agent project that helps builders handle neutral Browser Use for AI agents: real Chrome automation with a CLI.
Added
Safari Mcp
Safari Mcp is an MCP server that gives agents a way to handle native Safari browser automation for AI agents — 98 tools, zero Chrome.
Added
Qz Browser Agent
Qz Browser Agent is an open-source agent project that helps builders handle local-first AI agent for Chrome: browser automation, web research, and data extraction.
Added
FSB
FSB is an MCP server that gives agents a way to handle DOM-first AI browser automation for Chrome and MCP clients.
Added
BrowserStack MCP
BrowserStack's official MCP server, letting coding agents run manual and automated tests on real browsers and devices, debug failures and check accessibility from the IDE.
Added
GAL
GAL (General Agent Loop & Executor) is a self-contained agent runtime shipped as a single binary for Linux, macOS and Windows that you drive from a CLI, desktop app, browser, chat apps or HTTP, with a kernel-enforced sandbox and scoped one-time authorizations so every agent step is contained and auditable without Docker, Python or Node.
Added
Raydo Agent OS
The AI agent operating system for real work. Raydo gives autonomous agents a dedicated desktop workspace where they can browse the web, read and edit local files, run multi-step workflows across apps, and hand off tasks for human review. Built for founders, operators, and engineering teams that want background agents executing repeatable workflows with full step logs, sandboxing, and approval checkpoints.
Added
Anchor Action Index
Shared memory of web workflows for Anchor's cloud browsers: computer-use agents replay proven paths as deterministic code and the model only steps in where a page has changed, exposed as tools on Anchor's MCP server and REST API.
Added
ByteFlow
ByteFlow is a no-code builder for AI agents on top of the systems a company already runs (ERP, CRM, in-house and legacy apps) through APIs or a browser agent. Agents prepare the work, such as a CRM update or ERP entry, and a person approves it with one tap. Includes voice agents in 10+ Indian languages and document extraction.
Added
HARPA AI
Chrome extension browser agent that combines models like GPT, Claude and Gemini to automate web tasks, searching, writing and summarizing.
Added
LastBrowser
LastBrowser is a free, local-first AI browser for Windows, with a Linux preview, in beta: its Sidekick assistant helps find things and complete tasks, and power users get 48 curated agentic workflows, an MCP hub, isolated Spaces and a built-in PowerShell terminal, plus Widevine DRM for streaming.
Added
Ponta
Your AI app works on your Windows PCs: files, apps, browser and mail. Risky steps wait for your yes.
Added
Replit Agent
Builds and deploys full apps from a prompt in the browser.
Added
Resemble Agent Detection
Behavioral detection service from Resemble AI that tells websites and apps whether each session is a human, a classic bot or an AI agent such as ChatGPT agent, Claude in Chrome, Browser Use or Playwright automation. A tag-manager snippet feeds cursor movement, typing rhythm, clicks and scrolling to Resemble's System One model instead of trusting spoofable user-agent strings, and per-page rules decide what agents may do. Free account to start.
Added
SalesTouch
SalesTouch is a LinkedIn MCP server for AI agents. Connect Claude, Codex, ChatGPT or any remote MCP client to https://www.salestouch.io/api/mcp and your agent can find prospects, research people and companies, send invitations and messages, and handle follow-ups through 45 purpose-built LinkedIn tools instead of browser automation. Scoped OAuth, approvals for sensitive actions, daily limits, working hours and a human-paced action queue keep the account under control.
Added
GitHub repos (35)
| Repo | GitHub stars | Licence | Last push |
|---|---|---|---|
| 86,582 | Apache License 2.0 | 2026-10-26 | |
| 38,849 | Apache License 2.0 | 2026-10-22 | |
| 37,901 | GNU Affero General Public License v3.0 | 2026-10-30 | |
| 23,310 | GNU AGPL v3.0 | 2026-10-30 | |
| 20,809 | MIT | 2026-10-30 | |
| 16,629 | MIT | 2026-05-15 | |
| 16,584 | MIT | 2024-12-17 | |
| 14,204 | Apache License 2.0 | 2026-10-24 | |
| 13,236 | — | — | |
| 7,299 | — | — | |
| 5,682 | — | — | |
| 3,253 | MIT | 2026-08-14 | |
| 2,282 | — | — | |
| 2,222 | — | 2026-10-29 | |
| 2,148 | — | — |
Top 15 of 35 by stars. Stars and licence come from each repo, checked when the listing was last verified.
FAQ
What are the best AI tools for automating workflows focused on browser automation?
We list 43. The first three by GitHub stars and featured status are Crawl4AI: Crawl4AI is an Apache-2.0 web crawler that turns pages into clean, LLM-ready Markdown, extracts structured JSON and handles browser automation for RAG pipelines and agents, runnable yourself with Python or Docker. Crawl4AI Cloud offers the same capabilities as a hosted scrape, extract and search API with managed browsers and proxies, plus setup for Claude Code, Cursor and Codex.; Playwright MCP: Playwright MCP is Microsoft's Apache-2.0 MCP server that gives LLMs browser automation through Playwright, using structured accessibility snapshots instead of screenshots so no vision model is needed. It plugs into Claude, Cursor, VS Code, Codex and other MCP clients via npx @playwright/mcp; for high-throughput coding agents Microsoft now also offers the token-leaner Playwright CLI with skills.; Lightpanda: Lightpanda is an AGPL-3.0 headless browser written from scratch in Zig for AI agents, crawlers and test automation, not a Chromium fork. It executes JavaScript and exposes the Chrome DevTools Protocol, so existing Playwright and Puppeteer scripts can connect to it, and its own benchmark of 100 real pages shows about 16x less peak memory and 9x faster runs than headless Chrome. Install it via Homebrew or Docker, or use the hosted Lightpanda cloud.
Are there free or open-source options?
Yes: 28 open source, 1 source-available, 9 free or freemium, 3 paid, usage-based or enterprise, 2 pricing on the maker's site. Pricing is as each maker states it.
More from automating workflows
Tools are grouped by what each says it does. Missing one? Submit it free.