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
ego lite 对比 Playwright MCP

最佳 Playwright MCP 替代方案

Playwright MCP 是微软推出的浏览器自动化 MCP 服务器:它把导航、点击、snapshot 都封装成 MCP 工具,让任何支持 MCP 的 Agent 都能在不写测试脚本的情况下控制浏览器。

ego lite drives your real Chrome directly, batching multiple actions into one JavaScript call instead of one MCP tool call per step. Across 31 live browser automation jobs, ego lite finishes 96.8% of tasks with a perfect score; agent-driven Playwright finishes 71%.

开发者用户来自
GoogleAmazonShopifyTikTokHarvardStanfordUSCUCLA

Browser automation benchmark: ego lite vs Playwright MCP

This is agent-driven browser automation on live websites, not a snapshot-and-click loop on a fixture page. Both tools ran the same 31 multi-step tasks: several pages, several decisions, many behind a login. That is the work that turns one MCP tool call per action into a long, expensive chain. Both ran the identical set, so the bars below compare the same jobs.

agent: pimodel: ChatGPT 5.6 Solthinking: max31 tasks

Task completion rate

How often the agent finished the job. Stuck at login, skipped a step, or returned the wrong result: that task is a fail. Higher is better.

ego lite96.8%30 of 31 tasks perfect
playwright-cli71.0%22 of 31 tasks perfect

Cost per completed task

What one finished job costs in model spend. Failures still get billed, so they push this number up. Lower is better.

ego lite$1.98
playwright-cli$4.61

Model turns per task

How many times the model had to look at the page and choose the next action. Extra looks mean extra tokens, extra waiting, and extra places to stall. Lower is better.

ego lite30.3
playwright-cli42.8

Average task time

How long a job took from start to finish, on average, including time spent waiting on the model. Lower is better.

ego lite8m 38s
playwright-cli11m 33s

We ran every tool twice and kept the better score. A failed task still counts against it.

We measured Playwright through playwright-cli, not the MCP server. MCP uses the same engine and adds one extra round trip per action.

Check the numbers, or rerun the tasks yourselfThe 31 tasks, the grading checklists, and the raw results are open source. If a number looks off, open the repo.ego-browser-benchmark-framework

为什么 ego lite 比 Playwright MCP 更好

如果你已经在 Claude Code 或 Codex 里接了 Playwright MCP,那套循环你肯定很熟悉:snapshot、点击、等待、再 snapshot。每一步都要在模型里走一次往返。ego lite 用一个 Agent 原生浏览器取代了这套循环,你的编程 Agent 通过开源的 ego-browser 外壳来控制它。

96.8% of tasks done perfectly. No other tool tops 84%.

Playwright MCP re-reads the page through a tool call at every step, the browser sits idle while the model thinks, and every extra round trip is another chance for a long task to fall apart. On those 31 jobs, agent-driven Playwright ends 29% of tasks short of a perfect score.

In ego lite, every page reaches your agent as a Snapshot: a compressed accessibility tree with stable @N refs it can act on instantly, several actions per JavaScript turn. The agent spends its time doing the task, not parsing pages: 96.8% of the same 31 tasks end perfect, at the fastest average task time of the five tools measured.

Time to finish a task, shorter is better
518 sego (lite)
693 splaywright-cli
Data source - Task: finish an average live-site job

并行多任务,执行更快

一个 Playwright MCP 服务器只能驱动一个浏览器会话,想并行工作就得同时维护多套服务器配置。

在 ego lite 里,你在同一个浏览器里能开多少个 Space 没有上限:每个 Space 都在自己导入的 Chrome 配置文件上同时运行各自的任务,可以由同一个 Agent 驱动,也可以换成不同的 Agent。它们都不会抢占你正在用的标签页,你可以随时旁观或接管任意一个 Space。

不再反复来回 消耗更少 Token

Playwright MCP 的每一个动作都是一次独立的工具往返调用,都要把上下文重新过一遍模型。

在 ego lite 里,你的 Agent 每一轮只用几行 JavaScript 就能打包执行多个动作。叠加上 Snapshot 输入,同样的任务整体消耗的 Token 要少得多。

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 · Playwright MCP
>Add the first product to the cart, then check checkout loads
playwright - browser_navigate (MCP)(url: "localhost:3000")
Page Snapshot
… +224 lines (ctrl+o to expand)
1.6k tokens sent so far
Storefront rendered. Opening the first product.
playwright - browser_click (MCP)(ref: "e18")
Page Snapshot
… +198 lines (ctrl+o to expand)
3.1k tokens sent so far
Product page is up. Now the Add to cart button.
playwright - browser_click (MCP)(ref: "e42")
Page Snapshot
… +205 lines (ctrl+o to expand)
4.9k tokens sent so far
Added. Heading to checkout.
playwright - browser_navigate (MCP)(url: "localhost:3000/checkout")
Page Snapshot
… +231 lines (ctrl+o to expand)
6.8k tokens sent so far
Refs are stale after the navigation. Taking a fresh snapshot.
playwright - browser_snapshot (MCP)
Page Snapshot
… +238 lines (ctrl+o to expand)
9.2k tokens sent so far
Now waiting for the payment form to render.
playwright - browser_wait_for (MCP)(text: "Payment")
Page Snapshot
… +229 lines (ctrl+o to expand)
11.2k tokens sent so far
Confirming the form is actually there.
playwright - browser_evaluate (MCP)(expression: "!!document.querySelector('form#payment')")
Page Snapshot
… +25 lines (ctrl+o to expand)
13.6k tokens sent so far
─── 7 tool calls · 13.6k tokens ───
同一个任务,同一个模型。 左边:ego lite 一次批量 JavaScript 调用,总共 1.1k Token。 右边:Playwright MCP,7 次往返,Token 涨到 13.6k。

同一个 Chrome,为 Agent 而生

Playwright MCP 服务器启动的是一个空白浏览器,一遇到 SSO、2FA 和验证码就卡住。

ego lite 基于 Chromium 构建,一键导入你完整的 Chrome 环境。用起来和 Chrome 一模一样,你的浏览习惯完全不用改变。把它设为默认浏览器,你的 AI Agent 就再也不会卡在验证码、2FA 和 SSO 跳转上。

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 Playwright MCP

ego lite 与 Playwright MCP 的功能对比。
功能ego litePlaywright MCP
AI Agent 如何操作页面每一轮都写 JavaScript 把多个动作打包执行每个动作一次 MCP 工具调用,每次都要经过模型往返一趟
每个任务的 Token 消耗更低:Snapshot 输入加上批量操作,按整个任务计算更高:每一步都要在 MCP 循环里重新发送一次上下文
Browser automation on live sites (31 tasks, same model)96.8% perfect, $1.98 per completed task, 30.3 round trips71% perfect, $4.61 per completed task, 42.8 round trips (measured via playwright-cli)
复用你真实的 Chrome 登录状态可以,一键导入 Chrome 配置文件(Cookie、会话、扩展程序)不能。每次都启动全新浏览器,每个网站的登录都要手动配置
设置安装应用,在你的 Agent 中运行 /ego-browser在 Claude Code 或 Codex 各自的 MCP 配置中添加 Playwright MCP,然后自行管理服务器生命周期
并行任务隔离可以,Space 能在同一个浏览器里并排运行多个任务一台服务器一个会话,想并行就得开多个服务器实例
同时也是你的日常浏览器可以。你在自己的 Space 里浏览,Agent 在它们自己的 Space 里工作不能,仅限自动化专用的浏览器实例
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,无需订阅免费,开源
最近更新 2026年8月20日

让切换无缝衔接

已经在 Claude Code 或 Codex 里接好了 Playwright MCP?切换只需几分钟,而且不用卸载任何东西。

  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,检查页面有没有控制台报错

    把一句话粘贴进 Claude Code、Codex 或 Cursor,按下回车就行。不用配置 MCP 服务器,不用改 JSON,也不用写测试脚本。

  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

    任务在独立的 Space 中运行,不会影响你的标签页。浏览器窗口是真实存在的:可以实时查看,也可以随时接管。结果会返回到你的 AI Agent 的 CLI 中。

如果你愿意,CI 测试套件仍然可以继续用 Playwright MCP。ego lite 替代的是交互式的 Agent 浏览循环,不是你已经写好的 Playwright 测试代码。

各工具的适用场景

以下情况选择 ego (lite)

  • 你的任务要接触需要登录的网站:仪表盘、后台系统,或任何在 SSO、2FA 之后的东西。
  • 你想降低单个任务的 Token 账单。Snapshot 输入加上批量 JavaScript,效果胜过每个动作一次工具调用。
  • 你希望任务在并行的 Space 里运行,同时自己还能继续用同一个浏览器。
  • 你想彻底跳过 MCP 服务器配置:安装应用,运行 /ego-browser,就这么简单。

以下情况选择 Playwright MCP

  • 你已经深度使用 MCP 生态,希望在多个工具之间统一用同一套标准协议。
  • 你的自动化跑在 CI 或没有桌面浏览器的远程服务器上,需要无头模式。
  • 你需要覆盖 Firefox 或 WebKit。ego lite 只支持 Chromium。
  • 你要生成可提交的 Playwright 测试代码,希望 Agent 说的是同一套 API。

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。

还在权衡该选哪个?看看 Playwright MCP 和同类工具相比究竟如何。

常见问题

Playwright MCP 是微软为 Playwright 自动化框架推出的官方 MCP 服务器,以 playwright-mcp 的名义发布(你也会看到 mcp-playwright 或 playwright-mcp-server 这样的写法)。它把浏览器操作(导航、点击、输入、snapshot)都封装成 MCP 工具,让任何支持 MCP 的 Agent 都能在不写测试脚本的情况下操作浏览器。它读取页面时用的是无障碍树 snapshot 而不是截图,还附带一个 Chrome 扩展程序,也就是 Playwright MCP bridge,用来把服务器接到你正在运行的 Chrome 里的某个标签页上。它是“怎么给我的编程 Agent 配一个浏览器”这个问题的默认答案:维护良好、免费,还有 Playwright 团队背书。

在 Agent 驱动的浏览场景下是的。Playwright MCP 只是套在全新 Playwright 浏览器外面的一层轻量 MCP bridge;ego lite 则是一个完整的 Agent 浏览器:真实的 Chrome 配置文件、并行的 Space,还有能降低单任务 Token 消耗的批量 JavaScript 执行。如果你的目标是在 CI 里跑可复用、已固化的测试脚本,那 Playwright 本身仍然是更合适的工具。这方面的对比可以看我们的 ego lite 对比 Playwright 页面。

We measured the matchup through playwright-cli, the leaner of Microsoft's two agent interfaces. The MCP server shares the same Playwright engine and adds an MCP tool-call round trip on top of each action. On 31 live-site browser automation tasks, with the same pi agent, the same model (gpt-5.6-sol, thinking effort max), and the same written checklist, ego lite finished 30 of 31 tasks perfectly (96.8%) on 30.3 model turns at $1.98 per completed task; agent-driven Playwright finished 22 of 31 (71%) on 42.8 turns at $4.61. The tasks, the checklists, and the raw results are open source in the ego-browser-benchmark-framework repository on GitHub, so every number can be audited or reproduced.

常规做法是用 claude mcp add 把 Playwright MCP 添加到 Claude Code,然后一直保持 Playwright MCP 服务器运行。更快的做法是:安装 ego lite,在 Claude Code 里运行 /ego-browser。Agent 就能以 Snapshot 的方式读取页面,还能在浏览器里执行 JavaScript,不用配置 MCP 服务器,也不用改任何配置。

两者都支持。Codex 支持 MCP 服务器,所以 Playwright MCP 在 Codex 上的配置同样可行,用的是同一套“每个动作一次工具调用”的循环和空白浏览器配置文件。ego lite 接入 Codex 的方式和接入 Claude Code、Cursor 一样:运行 /ego-browser,Agent 就会驱动你真实登录的浏览器。Gemini CLI、OpenCode 以及自建 Agent 也都能用。

官方 Chrome 扩展程序能把 Playwright MCP 服务器接到你正在运行的 Chrome 里的某个标签页上,让 Agent 复用那个标签页的登录状态。这解决了单个标签页的空白配置文件问题,但循环本身没变(还是每个动作一次 MCP 工具调用),也没有隔离:Agent 和你共用同一个窗口工作。ego lite 会把你完整的配置文件和独立的 Space 都交给 Agent,任务并行运行,也不会碰到你的标签页。

Each browser action is a separate MCP tool call, and the model processes the result of every call before making the next one. Long tasks become long chains of round-trips, each carrying context. ego lite reduces the round-trips themselves: the agent batches several actions in one JavaScript execution and reads pages as compressed Snapshots, so the savings compound over a whole task. Across those 31 jobs, agent-driven Playwright averaged 42.8 model round trips per task to ego lite's 30.3, and that gap is most of the cost difference: $4.61 per completed task against $1.98.

开箱不行,因为它启动的是一个隔离的浏览器实例。你可以手动配置让它指向一个持久化配置文件或 CDP 端点,但遇到 SSO 和 2FA 时会话依然会失效。ego lite 只需导入一次你的 Chrome 配置文件,Agent 在任何地方都能继承真实的登录状态。

是的。ego lite 免费使用、无需订阅,ego-browser 外壳还是 MIT 协议的开源项目。