Este servidor MCP proporciona capacidades de automatización del navegador a través de Puppeteer, lo que permite la interacción tanto con nuevas instancias del navegador como con ventanas de Chrome existentes.
Este proyecto es una implementación experimental inspirada en @modelcontextprotocol/server-puppeteer . Si bien comparte objetivos y conceptos similares, explora enfoques alternativos para la automatización del navegador mediante el Protocolo de Contexto de Modelo.
Related MCP server: Puppeteer MCP Server
Navegar por páginas web
Tomar capturas de pantalla
Elementos de clic
Rellenar formularios
Seleccionar opciones
Elementos flotantes
Ejecutar JavaScript
Gestión inteligente de pestañas de Chrome:
Conectarse a pestañas activas de Chrome
Conservar las instancias de Chrome existentes
Manejo inteligente de conexiones
/
├── src/
│ ├── config/ # Configuration modules
│ ├── tools/ # Tool definitions and handlers
│ ├── browser/ # Browser connection management
│ ├── types/ # TypeScript type definitions
│ ├── resources/ # Resource handlers
│ └── server.ts # Server initialization
├── index.ts # Entry point
└── README.md # Documentationnpm install -g puppeteer-mcp-serverTambién puedes ejecutarlo directamente sin instalación usando npx:
npx puppeteer-mcp-serverClona este repositorio o descarga el código fuente
Instalar dependencias:
npm installConstruir el proyecto:
npm run buildEjecutar el servidor:
npm startPara utilizar esta herramienta con Claude, debes agregarla a tu archivo de configuración de MCP.
Agregue lo siguiente a su archivo de configuración de Claude Desktop (ubicado en %APPDATA%\Claude\claude_desktop_config.json en Windows o ~/Library/Application Support/Claude/claude_desktop_config.json en macOS):
{
"mcpServers": {
"puppeteer": {
"command": "puppeteer-mcp-server",
"args": [],
"env": {}
}
}
}{
"mcpServers": {
"puppeteer": {
"command": "npx",
"args": ["-y", "puppeteer-mcp-server"],
"env": {}
}
}
}{
"mcpServers": {
"puppeteer": {
"command": "node",
"args": ["path/to/puppeteer-mcp-server/dist/index.js"],
"env": {
"NODE_OPTIONS": "--experimental-modules"
}
}
}
}Agregue lo siguiente a su archivo de configuración MCP de la extensión Claude VSCode (ubicado en %APPDATA%\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json en Windows o ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json en macOS):
{
"mcpServers": {
"puppeteer": {
"command": "puppeteer-mcp-server",
"args": [],
"env": {}
}
}
}{
"mcpServers": {
"puppeteer": {
"command": "npx",
"args": ["-y", "puppeteer-mcp-server"],
"env": {}
}
}
}{
"mcpServers": {
"puppeteer": {
"command": "node",
"args": ["path/to/puppeteer-mcp-server/dist/index.js"],
"env": {
"NODE_OPTIONS": "--experimental-modules"
}
}
}
}Para la instalación de origen, reemplace path/to/puppeteer-mcp-server con la ruta real donde instaló esta herramienta.
El servidor lanzará una nueva instancia del navegador de forma predeterminada.
Para conectarse a una ventana de Chrome existente:
Cierre por completo cualquier instancia de Chrome existente
Inicie Chrome con la depuración remota habilitada:
# Windows
"C:\Program Files\Google\Chrome\Application\chrome.exe" --remote-debugging-port=9222
# macOS
/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --remote-debugging-port=9222
# Linux
google-chrome --remote-debugging-port=9222Navega a la página web deseada en Chrome
Conectarse usando la herramienta puppeteer_connect_active_tab :
{
"targetUrl": "https://example.com", // Optional: specific tab URL
"debugPort": 9222 // Optional: defaults to 9222
}El servidor hará lo siguiente:
Detectar y conectarse a la instancia de Chrome que se ejecuta con la depuración remota habilitada
Conserva tu instancia de Chrome (no la cerrará)
Buscar y conectarse a pestañas que no sean extensiones
Proporcionar mensajes de error claros si falla la conexión
Conéctese a una instancia de Chrome existente con la depuración remota habilitada.
Opcional:
targetUrl - URL de la pestaña específica a la que conectarse
debugPort : puerto de depuración de Chrome (predeterminado: 9222)
Navegar a una URL.
Obligatorio: url : la URL a la que navegar
Tome una captura de pantalla de la página actual o de un elemento específico.
Obligatorio: name - Nombre para la captura de pantalla
Opcional:
selector - Selector CSS para elemento a capturar
width - Ancho en píxeles (predeterminado: 800)
height - Altura en píxeles (predeterminado: 600)
Haga clic en un elemento de la página.
Obligatorio: selector - Selector CSS para el elemento en el que hacer clic
Rellene un campo de entrada.
Requerido:
selector - Selector CSS para campo de entrada
value - Texto para ingresar
Utilice menús desplegables.
Requerido:
selector - Selector CSS para el elemento seleccionado
value - Valor de la opción a seleccionar
Pase el cursor sobre los elementos.
Obligatorio: selector - Selector CSS para el elemento sobre el que se pasa el cursor
Ejecute JavaScript en la consola del navegador.
Requerido: script - Código JavaScript a ejecutar
Al utilizar la depuración remota:
Habilitar solo en redes confiables
Utilice un puerto de depuración único
Cerrar el puerto de depuración cuando no esté en uso
Nunca exponga el puerto de depuración a redes públicas
El servidor implementa un registro completo utilizando Winston:
Ubicación: logs/ directorio
Patrón de archivo: mcp-puppeteer-YYYY-MM-DD.log
Rotación de registro:
Rotación diaria
Tamaño máximo: 20 MB por archivo
Retención: 14 días
Compresión automática de registros antiguos
DEBUG: Información detallada de depuración
INFO: Información operativa general
ADVERTENCIA: Mensajes de advertencia
ERROR: Eventos de error y excepciones
Eventos de inicio/apagado del servidor
Operaciones del navegador (iniciar, conectar, cerrar)
Intentos de navegación y resultados
Ejecuciones y resultados de herramientas
Detalles de errores con seguimientos de pila
Salida de la consola del navegador
Uso de recursos (capturas de pantalla, registros de la consola)
El servidor proporciona mensajes de error detallados para:
Fallos de conexión
Elementos faltantes
Selectores no válidos
Errores de ejecución de JavaScript
Errores en la captura de pantalla
Cada llamada de herramienta devuelve:
Estado de éxito/fracaso
Mensaje de error detallado en caso de error
Datos del resultado de la operación en caso de éxito
Todos los errores también se registran en los archivos de registro con:
Marca de tiempo
Mensaje de error
Rastreo de pila (cuando esté disponible)
Información de contexto
¡Agradecemos sus contribuciones! Lea nuestras Pautas de Contribución para obtener más información sobre cómo enviar solicitudes de incorporación de cambios, informar problemas y contribuir al proyecto.
Este proyecto está licenciado bajo la licencia MIT: consulte el archivo de LICENCIA para obtener más detalles.
Click an element on the page
| Name | Required | Description | Default |
|---|---|---|---|
| selector | Yes | CSS selector for element to click |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description is a tautology of the name and adds no behavioral details beyond 'click'. It does not specify side effects (e.g., navigation trigger), waiting behavior, or error conditions. With no annotations, this is insufficient for safe invocation.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is very concise (one sentence) but lacks necessary detail. It is front-loaded but does not earn its place by providing substantive information beyond the tool name.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's simplicity (one parameter, no output schema), the description fails to cover essential context like post-click behavior, return value, or error handling. For a click action, crucial details (e.g., navigation wait) are missing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema already describes the selector parameter (CSS selector for element to click). The description adds no further meaning, but schema coverage is 100%, meeting the baseline. No extra value is provided.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('click') and the target ('an element on the page'), making the purpose obvious. It distinguishes from sibling tools like fill, hover, or navigate, which have different actions.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
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 (e.g., puppeteer_hover or puppeteer_fill). It does not mention prerequisites or context, leaving the agent to infer usage.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Connect to an existing Chrome instance with remote debugging enabled
| Name | Required | Description | Default |
|---|---|---|---|
| targetUrl | No | Optional URL of the target tab to connect to. If not provided, connects to the first available tab. | |
| debugPort | No | Optional Chrome debugging port (default: 9222) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, and the description does not disclose what happens if connection fails, whether state changes, or prerequisites (e.g., Chrome must already be running with remote debugging).
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single sentence with 10 words, efficiently conveying the core purpose without unnecessary details.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
With no output schema and no annotations, the description is too brief for a connection tool. It lacks information about return values, success/failure indicators, and prerequisites for using the tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Input schema coverage is 100% and already includes parameter descriptions. The tool description adds no additional behavioral details beyond what is in the schema, so baseline score of 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('Connect') and resource ('existing Chrome instance with remote debugging enabled'), distinguishing it from sibling tools that perform page manipulations.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool versus alternatives, such as launching a new browser. It does not indicate it is a prerequisite for other puppeteer actions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Execute JavaScript in the browser console
| Name | Required | Description | Default |
|---|---|---|---|
| script | Yes | JavaScript code to execute |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description bears full responsibility. It mentions execution in the browser console but does not specify whether the script runs synchronously, how return values are handled, or potential side effects on the page state.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single, direct sentence that immediately conveys the tool's purpose. It is front-loaded and contains no unnecessary words.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description is minimal given the tool's potential complexity. It does not explain what the execution returns, how to handle asynchronous code, or any security considerations. No output schema exists to compensate.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The single parameter 'script' has a clear description in the schema ('JavaScript code to execute'), and the tool description adds no extra meaning. With 100% schema coverage, baseline 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('Execute') and the target ('JavaScript in the browser console'), distinguishing it from sibling tools that perform specific UI interactions like clicking, filling, or navigating.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool versus alternatives such as puppeteer_click or puppeteer_fill. The description lacks context about typical use cases for executing arbitrary JavaScript.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Fill out an input field
| Name | Required | Description | Default |
|---|---|---|---|
| selector | Yes | CSS selector for input field | |
| value | Yes | Value to fill |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description does not disclose behavioral traits such as whether the field is cleared before filling, whether events are triggered, or what happens if the selector is invalid. With no annotations, the description carries the full burden but fails to provide any behavioral context.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single sentence that is front-loaded and concise. However, it could be slightly improved by adding more context without losing brevity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the simplicity of the tool (fill input field), the description is minimally adequate. However, the lack of output schema and behavioral details (e.g., does it clear existing values?) leaves gaps for the agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, with both 'selector' and 'value' having descriptions. The description adds no additional meaning beyond what the schema provides, so baseline 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Description states 'Fill out an input field' which clearly identifies the action (fill) and resource (input field). However, it does not differentiate from sibling tools like puppeteer_select or puppeteer_click, which could also interact with input fields.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
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 puppeteer_select or puppeteer_click. The description does not mention prerequisites, exclusions, or typical use cases.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Hover an element on the page
| Name | Required | Description | Default |
|---|---|---|---|
| selector | Yes | CSS selector for element to hover |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full burden. It only states the action without disclosing behavior like asynchronous execution, event side effects, or whether the hover is a one-time action.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single concise sentence. It is front-loaded and wastes no words, but could benefit from slightly more detail without losing conciseness.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the simplicity (1 param, no output schema), the description omits important context like return behavior (likely void) or side effects (e.g., triggering event listeners). Not complete for an agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% (selector described), so baseline is 3. The description adds no new meaning beyond the schema; it merely restates the action.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states the specific verb 'Hover' and resource 'an element on the page', clearly distinguishing it from sibling tools like click, fill, or navigate. However, it lacks nuance about what 'hover' entails, such as triggering mouseover events.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
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 puppeteer_click for interactions or puppeteer_evaluate for custom actions. There is no mention of prerequisites or context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Take a screenshot of the current page or a specific element
| Name | Required | Description | Default |
|---|---|---|---|
| name | Yes | Name for the screenshot | |
| selector | No | CSS selector for element to screenshot | |
| width | No | Width in pixels (default: 800) | |
| height | No | Height in pixels (default: 600) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full burden. It does not disclose what happens to the screenshot (e.g., saved to disk, returned as base64), nor any side effects or permissions needed.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single sentence, which is concise, but it lacks essential context such as output format or usage details. It is under-specified.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the lack of output schema and annotations, the description should explain what happens with the screenshot. It fails to mention return value or storage, making it incomplete.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, and the description adds no additional meaning beyond the schema. Baseline score of 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool takes a screenshot of the current page or a specific element, with a specific verb and resource. It differentiates from sibling tools like puppeteer_navigate or puppeteer_click.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool versus alternatives, no preconditions, and no mention of when to capture a full page versus an element.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Select an element on the page with Select tag
| Name | Required | Description | Default |
|---|---|---|---|
| selector | Yes | CSS selector for element to select | |
| value | Yes | Value to select |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description must carry the burden of behavioral disclosure. It only states the intended action but does not mention side effects (e.g., page state change), error conditions, or if it waits for the element.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single concise sentence. However, it is so brief that it sacrifices clarity; slightly more detail would be beneficial without losing conciseness.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple tool with well-described parameters, the description is adequate but incomplete. It does not explain change events, support for multiple select, or error handling, which would be useful for full completeness.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% with clear parameter descriptions. The tool description adds the context that it works with a 'Select tag', which clarifies the element type beyond the schema, but this is minimal additional value.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states 'Select an element on the page with Select tag', which identifies the target element type (select tag) and action, but it is vague about selecting an option from a dropdown. It does not clearly distinguish from sibling tools like puppeteer_click or puppeteer_fill.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
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 (e.g., puppeteer_click for generic clicks, puppeteer_fill for input fields). There is no mention of use cases or exclusions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
puppeteer_clickpuppeteer_connect_active_tabpuppeteer_evaluatepuppeteer_fillpuppeteer_hoverpuppeteer_navigatepuppeteer_screenshotpuppeteer_selectScored across 8 tools
Each tool targets a distinct browser action (click, navigate, fill, hover, etc.) with no overlap, making selection unambiguous.
All tools follow a consistent pattern: 'puppeteer_' prefix followed by a clear action verb, ensuring predictability.
With 8 tools covering essential browser interactions, the count is well-scoped for a focused automation server.
Core operations are covered, but missing explicit waiting or content extraction tools; however, puppeteer_evaluate can compensate.
AI-powered browser automation — navigate, click, fill forms, and extract data from any website.
Stealth web automation for AI agents. Login, signup, navigate, screenshot.
Stealth web automation for AI agents. Login, signup, navigate, screenshot.
Provides cloud browser automation capabilities using Stagehand and Browserbase, enabling LLMs to i…
