ego (lite) 只是一款浏览器；ego 才是你跨设备的个人 Agent。
加入候补名单
下载 Mac 版
解决方案
文档企业版更新日志博客
ego lite 对比 Selenium

最好用的 Selenium 替代品

Selenium is the W3C WebDriver standard for cross-browser tests: you write the script, tune the waits, and run it against a fresh browser with no logins.

ego lite is a Selenium alternative for AI agents. There is no script, no driver, and no Grid. Run /ego-browser and your coding agent controls ego lite, reading each page as a Snapshot and writing its own JavaScript, finishing whole tasks 3 to 4x faster than Selenium.

免费试用 ego (lite)GitHub
开发者用户来自
GoogleAmazonShopifyTikTokHarvardStanfordUSCUCLA

为什么 ego lite 比 Selenium 更好

Selenium still owns cross-browser regression in CI. Agent work inside the browser you already live in is a different job. Run /ego-browser in Claude Code, Codex, or Cursor and describe what you want.

真实任务中快 3 到 4 倍

Selenium's timing pain is familiar: waits you keep tuning, elements that go stale mid-click, retries baked into the starter config, and "Element is not clickable at point" forever on Stack Overflow.

Run /ego-browser and your coding agent controls ego lite, reading each page as a Snapshot with stable @N refs, acting on what's on the page, and reading again only if the page moved, finishing whole tasks 3 to 4x faster than Selenium.

Time to finish a task, shorter is better
81.8 sego lite
282.9 sagent browser
Data source - Task: scrape the main posts from the X account from the last 7 days

并行多任务，执行更快

Running two browser chores at once in Selenium means either juggling parallel driver sessions yourself or standing up Grid. That is the right machinery for a nightly cross-browser matrix, and far too much for three things you want done before lunch.

In ego lite there's no cap on how many Spaces you open inside the same browser: each one runs its own task at the same time, controlled by a coding agent via /ego-browser on its own imported Chrome profile. None of them ever steals the tab you're using, and you can watch or take over any Space anytime.

不再反复来回 消耗更少 Token

Raw Selenium scripts spend no tokens because there is no model in the loop. Wrap WebDriver for an agent and the bill shows up fast: Selenium still has no official MCP server, and community servers often dump a heavy DOM tree into the model's context.

Run /ego-browser and your coding agent batches several actions in a few lines of JavaScript per turn to control ego lite, reading pages as Snapshots of roughly 200 to 400 tokens. Stacked together, the same task completes on far fewer tokens overall.

Claude Code · ego lite
>/ego-browser Add the first product to the cart, then check checkout loads
●Bash(
ego-browser nodejs <<'EOF' await openOrReuseTab('localhost:3000') await click('.product-card') await click('.add-to-cart') await click('a[href="/checkout"]') cliLog('checkout loaded') EOF
)
⎿Output:
[t=4480ms] checkout loaded
Done. Both steps work.
─── 1 tool call · ↑ 1.1k tokens ───
Claude Code · mcp-selenium
>Add the first product to the cart, then check checkout loads
●selenium - start_browser (MCP)(browser: "chrome")
⎿Result
Browser started with session_id 4f1c…
↑ 1.6k tokens sent so far
Fresh Chrome, no cookies and no logins. Loading the storefront.
●selenium - navigate (MCP)(url: "localhost:3000")
⎿Result
Navigated to localhost:3000
↑ 3.1k tokens sent so far
Nothing came back about the page itself. Reading the accessibility resource to find the first product.
●selenium - accessibility://current
⎿Accessibility tree
… +214 lines (ctrl+o to expand)
↑ 4.9k tokens sent so far
First product card is .product-card. Clicking it.
●selenium - interact (MCP)(action: "click", by: "css", value: ".product-card")
⎿Result
Clicked element
↑ 6.8k tokens sent so far
Product page is up, so everything I just read is out of date. Reading the tree again.
●selenium - accessibility://current
⎿Accessibility tree
… +206 lines (ctrl+o to expand)
↑ 9.2k tokens sent so far
Add to cart is .add-to-cart.
●selenium - interact (MCP)(action: "click", by: "css", value: ".add-to-cart")
⎿Result
Clicked element
↑ 11.2k tokens sent so far
Added. Opening checkout, then I still have to confirm it loaded.
●selenium - interact (MCP)(action: "click", by: "css", value: "a[href='/checkout']")
⎿Result
Clicked element
↑ 13.6k tokens sent so far
─── 7 tool calls · ↑ 13.6k tokens ───
同样的任务，同样的模型。 左边：ego lite 里一次打包的 JavaScript 调用，总共 1.1k tokens。 右边：经社区 MCP 服务器的 Selenium，7 次往返，token 涨到 13.6k。

同一个 Chrome，为 Agent 而生

Selenium cannot attach to the Chrome profile you are already using. Point ChromeDriver at it while Chrome is open and you get "user data directory is already in use."

Built on Chromium, ego lite imports your entire Chrome setup in one click. When your coding agent runs /ego-browser, it controls ego lite with your real sessions already in place, so it doesn't stall on SSO, 2FA, or captchas on accounts you already have.

ego lite 的 Chrome 浏览器配置文件导入：一键设置，保留所有登录状态

ego lite 对比 Selenium

ego lite 与 Selenium 的功能对比。
功能ego liteSelenium
任务是怎么完成的Describe the task; coding agent controls ego lite via /ego-browser用 Java、Python、C#、Ruby 或 JavaScript 编写 WebDriver 脚本
等待与不稳定Coding agent reads Snapshots and acts via /ego-browserExplicit waits, stale-element handling, retry counts
复用你真实的 Chrome 登录状态Yes, one-click Chrome profile importNo. Fresh session per run; a live Chrome profile is locked
设置Install ego lite, run /ego-browser in your agentLanguage project plus a driver per browser
可与 AI Agent 配合使用Built for them via /ego-browserNo official MCP; community servers like mcp-selenium
并行任务Space 在一个可见的浏览器内隔离各个任务由你运维的 Grid 节点或并行 driver 会话
跨浏览器覆盖面仅 ChromiumChrome, Edge, Firefox, and Safari via WebDriver
日常使用的浏览器可以。你在自己的 Space 里浏览，Agent 在它们自己的 Space 里工作不是，它是面向测试的自动化库
可复用技能（即将上线）将成功的运行过程提炼为可复用技能；Agent 重复执行复杂任务时最高可快 5 倍（限量测试中）没有内置的同类功能
价格免费，无需订阅Free, Apache-2.0 (Grid capacity is your cost)
最近更新 2026年7月30日

让切换无缝衔接

Don't port your WebDriver suite. Move the interactive, logged-in work that never fit a regression suite to ego lite in a few minutes.

  1. 下载 ego (lite)

    下载 ego lite，一键导入你的 Chrome 浏览器配置文件，登录状态、Cookie 和扩展程序全部一起带过来。

  2. 用 /ego-browser 运行你的第一个任务

    粘贴到你的 Agent 里

    /ego-browser 打开 ego.app，列出页面上每个按钮的文字

    Paste one sentence into Claude Code, Codex, or Cursor and hit enter. No driver binary, no Grid, no MCP config.

  3. 看它开始工作
    ego lite 的 Space 概览，四个浏览器任务并排运行：Claude Code 在 Yahoo Finance 上追踪苹果股价，Codex 在 cars.com 上按年份筛选车型，Hermes 在完成一项 SaaS 后台任务，一位用户在抓取 X 上的数据，还有一只手正在点击 + 打开另一个 Space

    The task runs in its own Space, so your tabs stay untouched. Watch live or take over anytime. Results land back in your agent's CLI.

Keep Selenium for cross-browser regression in CI.

各工具的适用场景

以下情况选择 ego (lite)

  • Your tasks touch logged-in sites: dashboards, admin panels, anything behind SSO or 2FA.
  • Interactive, one-off browser work that never belonged in a regression suite.
  • You want several tasks running in parallel Spaces without standing up Grid.
  • You care about per-task token cost. Snapshot input plus batched JavaScript beats one MCP call per action.

以下情况选择 Selenium

  • You need Firefox and Safari coverage. ego lite is Chromium only.
  • Your team writes tests in Java, Python, C#, Ruby, or JavaScript with official WebDriver bindings.
  • You run enterprise regression on Selenium Grid.
  • You want the WebDriver standard, Apache-2.0 governance, and decades of QA answers on Stack Overflow.

为你的 Agent 配备真实浏览器

免费使用，运行在你的 Mac 上，一键导入 Chrome 浏览器配置文件。支持 Claude Code、Codex、Cursor，以及任何能写代码的 CLI Agent。

试用 ego (lite)

继续阅读

还在权衡？看看 Selenium 和同赛道其他工具怎么比。

ego (lite) vs Puppeteerego (lite) vs Playwright MCPego (lite) vs Browser Useego (lite) vs Browserbaseego (lite) vs Stagehandego (lite) vs agent-browserego (lite) vs Playwrightego (lite) vs Codex browser所有对比浏览器自动化解决方案

常见问题

Selenium is the industry's longest-running browser automation project and the reference WebDriver implementation, with bindings for Java, Python, C#, Ruby, and JavaScript plus Grid for distributed cross-browser runs. It is excellent for QA regression. Searches like Selenium vs Playwright or Selenium vs Puppeteer usually compare test frameworks; this page is about what changes when an AI agent, not a script, browses inside a browser you are already signed into.

Yes for agent-driven, logged-in browser work: describe the task, no WebDriver script, no driver binary. No for cross-browser regression suites or Selenium Grid pipelines in CI. Keep Selenium for committed tests; use ego lite for interactive work that never fit a suite.

Selenium was built for deterministic test scripts, not LLM-driven loops. Wrapping WebDriver for an agent means heavy DOM dumps or community MCP servers like mcp-selenium, with no official Selenium MCP from the project. ego lite is built for agents from the start: run /ego-browser and your coding agent controls ego lite with Snapshot input and batched JavaScript.

There is no official Selenium MCP server. Maintainers have noted that interactive LLM sessions saturate context quickly. Community servers like mcp-selenium exist but often re-read a heavy accessibility tree on every page change. ego lite skips MCP entirely: run /ego-browser in Claude Code, Codex, or Cursor.

Not the profile Chrome is already using. Selenium's issue tracker documents that ChromeDriver cannot attach to a live default profile while Chrome is open. ego lite imports a copy of your profile once, and every task starts signed in.

Playwright vs Selenium usually comes down to test engineering. Playwright is a modern framework with auto-waiting and strong debugging. Selenium is the W3C WebDriver standard with five official languages and Grid for cross-browser matrices. Both start from a blank browser. If an agent is doing the browsing, the bigger question is whether it can reuse your logins. See ego lite vs Playwright and ego lite vs Puppeteer.

For test automation, Selenium vs Puppeteer depends on language and browser reach. Puppeteer is a lean Node.js library focused on Chrome. Selenium is cross-browser and multi-language with Grid. For selenium puppeteer searches comparing AI agent paths, both still mean writing and maintaining scripts. ego lite removes that layer for logged-in agent work.

Neither is designed as an agent browser out of the box. Teams usually add Playwright MCP or a community Selenium MCP wrapper, which still means one tool call per action and a fresh or manually configured profile. ego lite is the agent-native path: run /ego-browser and your coding agent controls ego lite with fewer tokens per whole task.

navigator.webdriver is part of the WebDriver standard, so automated fingerprints are expected. ego lite is not a stealth tool. The practical difference is narrower: your coding agent controls ego lite, the Chromium browser you are already signed into, so many login walls never appear in the first place.

Free, local, no account. No ChromeDriver, GeckoDriver, or Grid to install or keep in sync. macOS 12+, Intel and Apple Silicon. Run /ego-browser in Claude Code, Codex, or Cursor.