ego (lite) はブラウザそのもの、ego は端末を横断して動くあなた専属のエージェントです。
順番待ちリストに参加する
Mac版をダウンロード
ソリューション
ドキュメントエンタープライズ更新履歴ブログ
ego lite vs Stagehand

最良の Stagehand 代替

Stagehand is Browserbase's open-source SDK: you write TypeScript or Python, and methods like act, extract, and observe turn plain-language instructions into browser actions.

ego lite is a Stagehand alternative that skips the SDK entirely. Run /ego-browser and your coding agent controls ego lite directly, batching a whole task into one JavaScript call, finishing whole tasks 3 to 4x faster than Stagehand.

ego (lite) を無料で試すGitHub
導入している開発者の所属先
GoogleAmazonShopifyTikTokHarvardStanfordUSCUCLA

ego lite が Stagehand より優れている理由

Stagehand means maintaining a TypeScript or Python project, and every uncached action is another model call at runtime. With ego lite you type one sentence in Claude Code, Codex, or Cursor after /ego-browser. No project, no schema, no API key.

実タスクで 3〜4倍速い

In Stagehand, three clicks in a row can mean three separate trips to the model before the page moves again. Stagehand's own speed guide shows this pattern and suggests calling observe first so later clicks can reuse what it found.

Run /ego-browser and your coding agent controls ego lite directly. Every page arrives as a Snapshot: a compressed accessibility tree with stable @N refs it can act on instantly. Your coding agent spends its time on the task, not restructuring scripts, finishing whole tasks 3 to 4x faster than Stagehand.

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

並列マルチタスクで、より速く実行

Running several Stagehand tasks at once means spinning up and tearing down browser instances in your own code, or moving to Browserbase and hitting session caps of 3, 25, or 100. To see what happened, you open a replay after the run finishes.

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.

何度も往復する必要はもうなし トークン消費も削減

Stagehand's action cache helps when you repeat the same flow on the same page. On a new site, or after the DOM shifts, each call asks the model again, and stale caches have to be cleared by hand.

Run /ego-browser and your coding agent batches several actions in a few lines of JavaScript per turn to control ego lite. Stacked with Snapshot input, 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 ───
Stagehand · env: "LOCAL"
>$ npx tsx run_task.ts const stagehand = new Stagehand({ env: "LOCAL" }) await stagehand.init()
await page.goto("http://localhost:3000")
// page = stagehand.context.pages()[0]
navigated: http://localhost:3000
↑ 1.6k tokens sent so far
await stagehand.act("click the first product card")
// LLM call #1 · model: gpt-5.6-sol
→ { success: true, message: "Action completed successfully",
actions: [{ selector: "/html/body/div[2]/main/div[1]/a",
method: "click", arguments: [] }] }
↑ 3.8k tokens sent so far
await stagehand.act("click add to cart")
// LLM call #2 · model: gpt-5.6-sol
→ { success: true, message: "Action completed successfully",
actions: [{ selector: "/html/body/div[2]/main/button[1]",
method: "click", arguments: [] }] }
↑ 5.9k tokens sent so far
await stagehand.observe("find the checkout link")
// LLM call #3 · model: gpt-5.6-sol
→ [{ selector: "/html/body/nav/a[href='/checkout']",
description: "Checkout link", method: "click", arguments: [] }]
↑ 8.0k tokens sent so far
await stagehand.act(action)
// no LLM: Action from observe()
→ { success: true, message: "Action completed successfully",
actions: [{ method: "click" }] }
navigated: /checkout
↑ 9.1k tokens sent so far
await stagehand.extract("is the payment form present?", z.object({ present: z.boolean() }))
// LLM call #4 · model: gpt-5.6-sol
→ { present: true }
↑ 11.4k tokens sent so far
await stagehand.extract("list the payment fields", z.object({ fields: z.array(z.string()) }))
// LLM call #5 · model: gpt-5.6-sol
→ { fields: ["card", "expiry", "cvc", "zip"] }
↑ 13.6k tokens sent so far
─── 7 model calls · ↑ 13.6k tokens ───
The same task, the same model. Left: one batched JavaScript call in ego lite, 1.1k tokens total. Right: a Stagehand script, a model call on nearly every step, tokens climb to 13.6k.

同じ Chrome を、エージェントネイティブに

Staying logged in is the hard part of local Stagehand. Chrome profile support still has an open bug in v3, cookie workarounds often miss the HttpOnly session cookies that matter, and the browse CLI only attaches if you started Chrome with remote debugging yourself.

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 の Chrome プロファイルインポート:ワンクリックで全ログイン情報をセットアップ

ego lite vs Stagehand

ego lite と Stagehand の機能比較。
機能ego liteStagehand
書くもの何もなし。すでに使っているエージェントへの一文A TypeScript or Python project: SDK, act/extract/observe, keep it working
エージェントがページをどう操作するかCoding agent batches multiple actions in JavaScript per turn via /ego-browserOne model call per act, extract, or observe that isn't cached; precise locators when you want to skip the model
タスクあたりのトークンコストLower: Snapshot plus batched JavaScript, measured per taskRepeated flows can reuse a cached selector; new pages pay for a fresh model call each time
実際のChromeログインをそのまま使い回すYes, one-click Chrome profile import; coding agent controls ego lite via /ego-browserNot reliably in local mode: Chrome profile support has an open bug, and browse needs remote debugging
セットアップInstall ego lite, run /ego-browser in your agentnpm/pip install, model API key, and a project
タスクを並行して分離Unlimited Spaces in one browser, bounded by your machineYour code manages instances; Browserbase caps concurrency by plan
普段使いのブラウザとしても利用可能はい。あなたは自分のSpaceで閲覧し、エージェントは自分たちのSpaceで作業しますいいえ。きれいな自動化用ブラウザを起動する
コミット可能で再現性のある自動化No. Tasks are interactiveYes: scripts with git history, review, and a committed cache
再利用可能なスキル (近日公開)成功した実行を再利用可能なスキルへと蒸留し、エージェントが繰り返すほど複雑なタスクで最大5倍高速化 (限定ベータ)Action caching on repeated flows only
価格無料、サブスクなし、アカウントなしMIT and free to install; you pay model tokens (plus Browserbase if hosted)
最終更新 2026年7月30日

移行をシームレスに

Already running Stagehand scripts in CI or on a schedule? Keep those. Move one-off and logged-in exploratory work to ego lite in a few minutes.

  1. ego (lite) をダウンロード

    ego liteをダウンロードして、ワンクリックでChromeプロファイルを取り込みます。ログイン情報もCookieも拡張機能もそのまま引き継がれます。

  2. /ego-browserで最初のタスクを実行

    エージェントに貼り付ける

    /ego-browser ego.app を開いてウェイトリストに参加

    Paste one sentence into Claude Code, Codex, or Cursor and hit enter. No SDK, no model API key, no Zod schema.

  3. 動作を見る
    4つのブラウザタスクを並べて表示する ego lite Space の概要:Claude Code が Yahoo Finance で Apple 株を追跡し、Codex が cars.com で年式ごとに車をフィルタリングし、Hermes が SaaS のバックオフィス業務を仕上げ、ユーザーが X をスクレイピングし、手が + をタップして新しい Space を開いている

    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 Stagehand for committed automation, typed extraction, and deterministic locators.

それぞれのツールを使うべき場面

ego (lite) を選ぶべきケース

  • You'd rather describe a one-off scrape or form fill than maintain a TypeScript or Python project for it.
  • The work sits behind your own logins and should start signed in.
  • You want parallel Spaces you can watch, without managing browser instances in code.
  • You don't want to add an SDK, model key, and project for something your coding agent can already do.

Stagehand を選ぶべき場合

  • Automation must live in a repo with review, caching, and the same run months later.
  • You want a deterministic escape hatch: precise locators when a flow must not vary.
  • You need typed extraction with a Zod schema into TypeScript.
  • It must run in CI or across many cloud browsers on Browserbase.

AI エージェントに本物のブラウザを

無料で Mac 上で動作し、ワンクリックで Chrome のプロファイルをインポートできます。Claude Code、Codex、Cursor、コードを書くあらゆる CLI エージェントに対応しています。

ego (lite) を試す

続きを読む

まだ比較中ですか？Stagehand が同分野の他ツールとどう違うかを見てください。

ego (lite) vs Puppeteerego (lite) vs Playwright MCPego (lite) vs Browser Useego (lite) vs Browserbaseego (lite) vs agent-browserego (lite) vs Playwrightego (lite) vs Seleniumego (lite) vs Codex browserすべての比較ブラウザ自動化ソリューション

よくある質問

Stagehand is Browserbase's MIT-licensed browser automation SDK for TypeScript and Python. It gives you act, extract, observe, and agent primitives, plus precise locators when you want to skip the model. It runs locally against Chromium, and pairs with Browserbase when you want hosted sessions, caching, and replay. For automation you commit to a repo and run on a schedule, it remains an excellent choice.

For interactive, logged-in browser work, yes: run /ego-browser and describe the task. For committed automation you ship to CI, with typed extraction and deterministic locators, keep Stagehand. Stagehand is for browser work you commit; ego lite is for browser work you do today.

No. Stagehand runs locally against Chromium on its own. Browserbase is optional when you want hosted sessions, server-side caching, and replay. The Browserbase vs Stagehand pairing matters for cloud fleets, not for a local script on your laptop. ego lite has no hosted path: local only, free, no account.

Often yes in production, but for different reasons. Stagehand is the SDK you write act, extract, and observe calls in. Browserbase is where those scripts run when you need cloud scale. For your own agent chores on logged-in sites, you may need neither: run /ego-browser and let Claude Code or Codex drive ego lite directly.

Stagehand is MIT-licensed and free to install. You pay model tokens for every act, extract, or observe call that is not cached, plus Browserbase browser minutes if you host it there. ego lite is free. You only pay the model tokens your coding agent already uses.

Not dependably in local mode today. Pointing Stagehand at your Chrome profile still has an open bug in v3, and cookie workarounds often miss the HttpOnly session cookies that matter. ego lite imports your Chrome profile once, and every task starts signed in.

They aim at different developers. Stagehand is TypeScript-first with act, extract, and observe primitives for hybrid code-plus-AI automation. Browser Use is Python-first and runs a more autonomous agent loop. Both are frameworks you write and fund. ego lite is the Chromium browser your coding agent controls via /ego-browser. See ego lite vs Browser Use for that side.

Stagehand is an SDK: you write act, extract, and observe calls in TypeScript or Python. agent-browser is a CLI your agent shells out to, one command at a time. ego lite is neither. Run /ego-browser and your coding agent controls ego lite with batched JavaScript. See ego lite vs agent-browser.

Not as a native skill. Claude Code would run a Stagehand script you wrote, with model API keys and a project to maintain. ego lite is the shorter path: run /ego-browser in Claude Code, Codex, or Cursor and describe the task. No SDK install, no schema, no separate model billing loop.

When an action is not cached, the model picks the target again each time, so a shifting page can resolve differently. Caches can also go stale after the DOM changes. Use Stagehand's precise locators when a flow must not vary. ego lite is built for interactive work where some variation is acceptable.

Neither. Run /ego-browser in Claude Code, Codex, Cursor, Gemini CLI, or OpenCode. Free on macOS 12+, no account required.