Skip to main content
Glama
enesjakozh
ztobs

Browser Use Server

by ztobs
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Hybrid

ブラウザ使用サーバー

鍛冶屋のバッジ

Pythonスクリプトを使用したブラウザ自動化のためのモデルコンテキストプロトコルサーバー。Clineで使用

特徴

ブラウザ操作

  • screenshot : ウェブページ（全ページまたはビューポート）のスクリーンショットをキャプチャします

  • get_html : ウェブページのHTMLコンテンツを取得する

  • execute_js : ウェブページでJavaScriptを実行する

  • get_console_logs : Webページからコンソールログを取得する

すべての操作は、ページの読み込み後のカスタム インタラクション ステップ (要素のクリック、スクロールなど) をサポートします。

Related MCP server: Playwright MCP Server for Security

前提条件

  1. (オプションですが推奨) ヘッドレス ブラウザ自動化用に Xvfb をインストールします。

# Ubuntu/Debian
sudo apt-get install xvfb

# CentOS/RHEL
sudo yum install xorg-x11-server-Xvfb

# Arch Linux
sudo pacman -S xorg-server-xvfb

Xvfb（X Virtual Frame Buffer）は仮想ディスプレイを作成し、ボットとして検知されることなくブラウザの自動化を可能にします。Xvfbの詳細については、こちらをご覧ください。

  1. MinicondaまたはAnacondaをインストールする

  2. Conda 環境を作成します。

conda create -n browser-use python=3.11
conda activate browser-use
pip install -r requirements.txt

  1. LLM 構成をセットアップします。

サーバーは複数のLLMプロバイダーをサポートしています。以下のAPIキーのいずれかを使用できます。

# Required: Set at least one of these API keys
export GLHF_API_KEY=your_api_key
export GROQ_API_KEY=your_api_key
export OPENAI_API_KEY=your_api_key
export OPENROUTER_API_KEY=your_api_key
export GITHUB_API_KEY=your_api_key
export DEEPSEEK_API_KEY=your_api_key
export GEMINI_API_KEY=your_api_key
export OLLAMA_API_KEY=your_api_key

# Optional: Override default configuration
export MODEL=your_preferred_model  # Override the default model
export BASE_URL=your_custom_url    # Override the default API endpoint
export USE_VISION=false  # Enable/disable vision capabilities (default: false)

サーバーは最初に見つかった利用可能なAPIキーを自動的に使用します。オプションで、環境変数を使用して、プロバイダーのモデルとベースURLをカスタマイズできます。

インストール

Smithery経由でインストール

Smithery経由で Claude Desktop 用の Browser Use Server を自動的にインストールするには:

npx -y @smithery/cli install @ztobs/cline-browser-use-mcp --client claude

  1. このリポジトリを/home/YOUR_HOME/Documents/Cline/ディレクトリにクローンします。

  2. 依存関係をインストールします:

npm install

  1. サーバーを構築します。

npm run build

MCP構成

Cline MCP 設定に次の構成を追加します。

"browser-use": {
  "command": "node",
  "args": [
    "/home/YOUR_HOME/Documents/Cline/MCP/browser-use-server/build/index.js"
  ],
  "env": {
    // Required: Set at least one API key
    "GLHF_API_KEY": "your_api_key",
    "GROQ_API_KEY": "your_api_key",
    "OPENAI_API_KEY": "your_api_key",
    "OPENROUTER_API_KEY": "your_api_key",
    "GITHUB_API_KEY": "your_api_key",
    "DEEPSEEK_API_KEY": "your_api_key",
    "GEMINI_API_KEY": "your_api_key",
    "OLLAMA_API_KEY": "your_api_key",
    // Optional: Configuration overrides
    "MODEL": "your_preferred_model",
    "BASE_URL": "your_custom_url",
    "USE_VISION": "false"
  },
  "disabled": false,
  "autoApprove": []
}

交換する：

  • YOUR_HOME実際のホームディレクトリ名に置き換えます

  • your_api_key実際の API キーに置き換えます

使用法

サーバーを実行します。

node build/index.js

サーバーは stdio で利用可能になり、次の操作をサポートします。

スクリーンショット

パラメータ:

  • url: ウェブページのURL（必須）

  • full_page: ページ全体をキャプチャするか、ビューポートのみをキャプチャするか（オプション、デフォルト: false）

  • 手順: ページの読み込み後に実行する手順をカンマで区切って記述するアクションまたは文章（オプション）

HTMLを取得

パラメータ:

  • url: ウェブページのURL（必須）

  • 手順: ページの読み込み後に実行する手順をカンマで区切って記述するアクションまたは文章（オプション）

JavaScriptを実行する

パラメータ:

  • url: ウェブページのURL（必須）

  • スクリプト: 実行する JavaScript コード (必須)

  • 手順: ページの読み込み後に実行する手順をカンマで区切って記述するアクションまたは文章（オプション）

コンソールログを取得する

パラメータ:

  • url: ウェブページのURL（必須）

  • 手順: ページの読み込み後に実行する手順をカンマで区切って記述するアクションまたは文章（オプション）

クラインの使用例

Cline でブラウザ用サーバーを使用して実行できるタスクの例を次に示します。

開発中のWebページ要素の変更

認証が必要なページの見出しの色を変更するには:

Change the colour of the headline with the text "Alle Foren im Überblick." to deep blue on https://localhost:3000/foren/ page

To check/see the page, use browser-use MCP server to:
Open https://localhost:3000/auth,
Login with ztobs:Password123,
Navigate to https://localhost:3000/foren/,
Accept cookies if required

hint: execute all browser actions in one command with multiple comma-separated steps

このタスクでは次のことを示します。

  • カンマ区切りのステップを使用した複数ステップのブラウザ自動化

  • 認証処理

  • クッキーの承認

  • DOM操作

  • CSSスタイルの変更

サーバーはこれらのステップを順番に実行し、その途中で必要なやり取りを処理します。

構成

LLM 構成

サーバーは、デフォルト構成で複数の LLM プロバイダーをサポートします。

  • GLHF: deepseek-ai/DeepSeek-V3 モデルを使用

  • Ollama: 32k コンテキスト ウィンドウの qwen2.5:32b-instruct-q4_K_M モデルを使用します

  • Groq: deepseek-r1-distill-llama-70b モデルを使用

  • OpenAI: gpt-4o-mini モデルを使用

  • Openrouter: deepseek/deepseek-chat モデルを使用

  • Github: gpt-4o-mini モデルを使用

  • DeepSeek: deepseek-chat モデルを使用

  • Gemini: gemini-2.0-flash-exp モデルを使用

環境変数を使用してこれらのデフォルトを上書きできます。

  • MODEL : 任意のプロバイダーのカスタムモデル名を設定します

  • BASE_URL : カスタム API エンドポイント URL を設定します (プロバイダーがサポートしている場合)

視力サポート

サーバーは、USE_VISION 環境変数を通じてビジョン機能をサポートします。

  • ブラウザ操作のビジョン機能を有効にするには、USE_VISION=true を設定します。

  • ビジョンが必要ない場合にパフォーマンスを最適化するために、デフォルトは false です。

  • ウェブページのコンテンツを視覚的に理解する必要があるタスクに役立ちます

Xvfb サポート

サーバーは、Xvfb がインストールされているかどうかを自動的に検出し、次の操作を実行します。

  • 利用可能な場合は xvfb-run を使用し、ボット検出なしでブラウザの自動化を改善します。

  • Xvfbがインストールされていない場合は直接実行にフォールバックします

  • RUNNING_UNDER_XVFB環境変数を適宜設定します

タイムアウト

デフォルトのタイムアウトは5分（300000ミリ秒）です。これを変更するには、 build/index.jsのTIMEOUT定数を変更してください。

エラー処理

サーバーは、次の詳細なエラー メッセージを提供します。

  • Pythonスクリプトの実行失敗

  • ブラウザ操作のタイムアウト

  • 無効なパラメータ

デバッグ

デバッグには MCP インスペクタを使用します。

npm run inspector

用途

ブラウザの使用

ライセンス

マサチューセッツ工科大学

Available Tools

4 tools
execute_jsC

Execute JavaScript code on a webpage

ParametersJSON Schema
NameRequiredDescriptionDefault
urlYesThe URL to navigate to
scriptYesThe JavaScript code to execute
stepsNoComma-separated actions or sentences describing steps to take after page load (e.g., "click #submit, scroll down" or "Fill the login form and submit")

TDQS

C2.9/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden of behavioral disclosure. It mentions execution but lacks details on permissions needed, potential side effects (e.g., page modifications), error handling, or execution environment. This is inadequate for a tool that performs code execution.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, clear sentence with no wasted words. It's front-loaded and efficiently communicates the core function without unnecessary elaboration.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness2/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the complexity of executing JavaScript on a webpage, the lack of annotations, and no output schema, the description is insufficient. It doesn't cover behavioral aspects like safety, return values, or error conditions, leaving significant gaps for an agent to use this tool effectively.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the schema already documents all parameters (url, script, steps). The description adds no additional meaning or context beyond what's in the schema, such as examples or constraints, but doesn't contradict it either.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the action ('Execute JavaScript code') and target ('on a webpage'), which is specific and unambiguous. However, it doesn't explicitly differentiate from sibling tools like get_console_logs or get_html, which also interact with webpages but for different purposes.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides no guidance on when to use this tool versus alternatives like get_html or screenshot, nor does it mention prerequisites or constraints. It simply states what the tool does without context for selection.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

get_console_logsC

Get the console logs of a webpage

ParametersJSON Schema
NameRequiredDescriptionDefault
urlYesThe URL to navigate to
stepsNoComma-separated actions or sentences describing steps to take after page load (e.g., "click #submit, scroll down" or "Fill the login form and submit")

TDQS

C2.9/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden of behavioral disclosure. It states what the tool does but lacks critical details: it doesn't specify if this requires browser automation, what types of console logs are captured (e.g., errors, warnings), whether it's a read-only operation, or any limitations like timeouts or authentication needs. This leaves significant gaps in understanding how the tool behaves.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, clear sentence with zero waste—it directly states the tool's purpose without unnecessary words. It's appropriately sized and front-loaded, making it easy to grasp immediately.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness2/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the complexity of interacting with webpages and the lack of annotations and output schema, the description is incomplete. It doesn't address key contextual aspects like what the tool returns (e.g., log format, error handling), behavioral constraints, or how it differs from siblings. For a tool with two parameters and no structured safety hints, more detail is needed to be fully helpful.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The input schema has 100% description coverage, clearly documenting both parameters ('url' and 'steps'). The description adds no additional meaning beyond what the schema provides, such as explaining the format of console logs or how steps interact with log capture. Since the schema does the heavy lifting, the baseline score of 3 is appropriate.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's purpose with a specific verb ('Get') and resource ('console logs of a webpage'), making it immediately understandable. However, it doesn't differentiate from sibling tools like 'execute_js' or 'get_html', which might also interact with webpage content, so it doesn't reach the highest score.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides no guidance on when to use this tool versus alternatives like 'execute_js' (which might execute JavaScript and potentially capture logs) or 'get_html' (which retrieves HTML content). There's no mention of prerequisites, such as whether the webpage needs to be loaded first, or any exclusions.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

get_htmlC

Get the HTML content of a webpage

ParametersJSON Schema
NameRequiredDescriptionDefault
urlYesThe URL to navigate to
stepsNoComma-separated actions or sentences describing steps to take after page load (e.g., "click #submit, scroll down" or "Fill the login form and submit")

TDQS

C2.9/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden of behavioral disclosure. It states what the tool does but doesn't describe how it behaves—e.g., whether it follows redirects, handles authentication, respects rate limits, or returns errors. This leaves critical operational details unspecified.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, clear sentence with zero wasted words. It's front-loaded and efficiently communicates the core function without unnecessary elaboration, making it easy to parse quickly.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness2/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the lack of annotations and output schema, the description is incomplete for a tool with two parameters and potential behavioral complexity. It doesn't address what the tool returns (e.g., raw HTML, status codes), error handling, or dependencies, leaving significant gaps for an AI agent to infer.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the schema already documents both parameters ('url' and 'steps') thoroughly. The description doesn't add any meaning beyond what the schema provides, such as clarifying the interaction between parameters or providing examples of 'steps' usage, resulting in a baseline score.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the action ('Get') and resource ('HTML content of a webpage'), making the purpose immediately understandable. It doesn't specifically differentiate from sibling tools like 'execute_js' or 'screenshot', but the core function is unambiguous.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description provides no guidance on when to use this tool versus alternatives like 'execute_js' or 'screenshot'. It doesn't mention prerequisites, limitations, or scenarios where this tool is preferred over others, leaving usage context entirely implicit.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

screenshotC

Take a screenshot of a webpage

ParametersJSON Schema
NameRequiredDescriptionDefault
urlYesThe URL to navigate to
full_pageNoWhether to capture the full page or just the viewport
stepsNoComma-separated actions or sentences describing steps to take after page load (e.g., "click #submit, scroll down" or "Fill the login form and submit")

TDQS

C2.9/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries full burden but only states the basic action without disclosing behavioral traits. It lacks details on permissions needed, potential rate limits, output format (e.g., image type), error handling, or whether it's a read-only or mutative operation, leaving significant gaps for an agent.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, efficient sentence with zero waste, front-loading the core purpose. Every word earns its place, making it highly concise and well-structured for quick comprehension.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness2/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity (involving webpage interaction and screenshot capture), no annotations, and no output schema, the description is incomplete. It fails to address critical context like what the output returns (e.g., image data or file path), error conditions, or behavioral nuances, leaving the agent under-informed.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the schema already documents all parameters (url, full_page, steps). The description adds no additional meaning beyond implying webpage capture, which is redundant with the schema's details. Baseline 3 is appropriate as the schema does the heavy lifting.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the verb ('Take') and resource ('screenshot of a webpage'), making the purpose immediately understandable. It distinguishes from siblings like execute_js or get_html by focusing on visual capture rather than code execution or HTML retrieval, though it doesn't explicitly name alternatives.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

No guidance is provided on when to use this tool versus alternatives like get_html for content extraction or execute_js for interactive actions. The description implies usage for webpage capture but offers no context about prerequisites, limitations, or comparative scenarios with sibling tools.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 4 tool updates
    • First observedexecute_js
    • First observedget_console_logs
    • First observedget_html
    • First observedscreenshot
Deploy Server
Give agents eyes on the webHosted screenshot MCP. Send a URL, get a pixel-perfect image. $0.002 per grab. Works with Claude, Cursor, and Codex.grabbit.liveAd

TDQS

A3.5/5.0

Scored across 4 tools

Disambiguation5/5

Each tool has a clearly distinct purpose: execute_js runs code, get_console_logs retrieves logs, get_html fetches content, and screenshot captures visual output. There is no overlap or ambiguity between these functions.

Naming Consistency5/5

All tool names follow a consistent verb_noun pattern with snake_case (e.g., execute_js, get_console_logs, get_html, screenshot). The naming is predictable and readable throughout.

Tool Count5/5

With 4 tools, this server is well-scoped for browser automation, covering key operations like executing scripts, retrieving logs, getting content, and taking screenshots. Each tool earns its place without being excessive or insufficient.

Completeness4/5

The tool set covers essential browser interactions for the domain, including execution, logging, content retrieval, and visualization. A minor gap exists in navigation or page manipulation tools (e.g., navigate, click), but core workflows are well-supported.

Maintenance

ActivityInactive
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

Appeared in Searches