Playwright-style browser API
Use familiar methods including `goto`, `click`, `locator`, and `screenshot`, making the SDK recognizable to developers who already work with Playwright-style automation.
Stagehand
Freemium
Stagehand is an SDK for building browser agents with Playwright-style APIs, natural-language actions, structured extraction, and self-healing behavior. It supports TypeScript, Python, and Go for agents that need to operate web pages with browser-level control.
Stagehandとは？
Stagehand is an open-source SDK for building browser agents. It provides Playwright-style browser control alongside natural-language primitives such as act, observe, and extract, so an agent can navigate pages, carry out tasks, and return structured information.
The SDK is available for TypeScript, Python, and Go. Its documented workflow combines normal browser operations such as goto, click, locator, and screenshot with agent-directed actions and schema-based extraction. Stagehand is intended for browser automation where model flexibility needs to be combined with control over the page and the browser runtime.
Stagehandでできること
Natural-language browser actions
The `act`, `observe`, and `extract` primitives let agents describe actions and page observations in natural language rather than relying only on fixed selectors.
Self-healing actions
When a site changes, Stagehand can detect selector changes and refresh how an action is performed, helping existing browser tasks recover from page updates.
Structured data extraction
Extract page information against a schema, as shown in the TypeScript example that returns story titles and point totals as structured results.
Agent-focused browser context
The SDK uses accessibility-tree trimming to provide agents with more targeted page context and reduce unnecessary information during browser interaction.
Support for complex browser features
The site lists support for nested iframes, closed Shadow DOM, WebMCP, clipboard operations, batched commands, and OpenTelemetry support.
利用シーン
“Natural-language page automation”
Build agents that navigate a website and carry out a stated task, such as upvoting the top story on a news site, while retaining direct browser control in code.
“Structured web research”
Collect selected page data into a defined schema—for example, extracting the top five stories and their point totals from a news page.
“Order-status workflows”
Automate browser interactions for tasks such as checking an order’s shipping status, an example highlighted in the project’s introductory material.
“Browser-agent benchmarking”
Use the project’s evaluation work as a reference point for assessing browser agents by accuracy, speed, and cost, or run evaluations with a chosen model, harness, and benchmark.
よくある質問
What is Stagehand used for?
Stagehand is used to build browser agents that navigate web pages, perform actions described in natural language, observe page content, and extract structured data.
Which programming languages does Stagehand support?
The site presents Stagehand for TypeScript, Python, and Go. The documented quickstart uses TypeScript.
How does a typical Stagehand workflow work?
A typical workflow creates a Stagehand instance, gets the active page, navigates with a browser method such as `goto`, performs an action with `act`, extracts data with a schema, and then closes the session.
Can Stagehand return structured data?
Yes. The documented example uses `extract` with a schema and returns typed story data, including titles and points.
What browser structures and capabilities are supported?
The site lists support for nested or out-of-process iframes, closed Shadow DOM, WebMCP, clipboard operations, batched commands, and OpenTelemetry support.
クイック情報
- Category
- Developer Tool
- Product type
- Open-source browser-agent SDK
- Languages
- TypeScript, Python, and Go
- Core primitives
- Act, Observe, and Extract
- Example output
- Schema-based structured page data
- Source domain
- stagehand.dev
Stagehandの代替品
Simular
www.simular.ai
Simular is an autonomous computer platform that uses AI agents to operate real computers and automate digital workflows. It serves individuals, builders, operators, and businesses that need to automate tasks across desktop applications and websites, including workflows that do not expose traditional APIs.
Yutori
yutori.com
Yutoriは、ブラウザ自動化やウェブ調査、定期的なウェブ監視に対応するAIエージェントを提供し、反復的なデジタル業務をクラウドエージェントとブラウザ操作向けモデルに任せられます。
Context
context.ai
Context は、顧客インフラ上で AI エージェントを構築・デプロイ・改善できる企業向けプラットフォームです。ワークスペース、ランタイム、コンテキスト、評価ツール、コネクタ、IdP ベースのアクセス制御を備えています。
Gemini Enterprise Agent Platform
cloud.google.com
Gemini Enterprise Agent Platform, formerly Vertex AI, is Google Cloud’s platform for developers and technical teams to build, deploy, govern, and optimize AI agents, generative AI applications, and machine learning models.
OneReach.ai
onereach.ai
OneReach.ai GSXは、企業向けAIエージェントを大規模に構築・運用・統制する基盤です。
Octomind Cloud
octomind.run
Octomind Cloud runs AI agents on persistent cloud machines for research, writing, web work, software development, and scheduled routines. Users give instructions in plain language and can receive results through the web panel, Telegram, WhatsApp, or Slack.