de en es ja ko ru zh

agent-browser

Name: agent-browser
Author: Nirmalraj10567

by Nirmalraj10567

Overview Schema Source Related Servers Discussions Admin

TypeScript

Hybrid

Agent Browser MCP Server

AI 어시스턴트를 위한 브라우저 자동화 도구를 제공하기 위해 agent-browser를 래핑하는 MCP(Model Context Protocol) 서버입니다. stdio 및 HTTP/SSE 전송을 모두 지원합니다.

기능

다음을 포함한 40개 이상의 브라우저 자동화 도구:

카테고리	도구
탐색	`browser_open`, `browser_navigate`, `browser_close`
상호작용	`browser_click`, `browser_fill`, `browser_type`, `browser_press`, `browser_hover`, `browser_select`, `browser_scroll`
검사	`browser_snapshot`, `browser_screenshot`, `browser_read`, `browser_get`, `browser_find`
실행	`browser_eval`, `browser_batch`, `browser_wait`
상태	`browser_cookie`, `browser_storage`, `browser_state`, `browser_session`, `browser_tab`
DevTools	`browser_console`, `browser_trace`, `browser_network`, `browser_react`, `browser_vitals`, `browser_a11y`
고급	`browser_frame`, `browser_stream`, `browser_dialog`, `browser_diff`, `browser_chat`, `browser_profile`
시스템	`browser_install`, `browser_doctor`, `browser_skills`, `browser_set`

Related MCP server: selenium-mcp

설치

npm install
npm run build

사용법

Stdio 모드(로컬 MCP 클라이언트용)

npm run dev

HTTP/SSE 모드(원격 액세스용)

node server.js

서버는 기본적으로 포트 5888에서 수신 대기합니다.

환경 변수

변수	기본값	설명
`PORT`	`5888`	서버 포트
`AGENT_BROWSER_SCREENSHOT_DIR`	`./screenshots`	저장된 스크린샷 디렉터리
`BASE_URL`	`/screenshots`	스크린샷의 공개 URL 경로
`MCP_SCRIPT_PATH`	`./dist/index.js`	컴파일된 MCP 서버 경로
`ENABLE_HTTPS`	`false`	Let's Encrypt 인증서로 HTTPS 활성화
`CERT_PATH`	`/etc/letsencrypt/live/test2.infinitrocyber.com`	SSL 인증서 경로

Docker

docker build -t agentbrowser-mcp .
docker run -p 5888:5888 agentbrowser-mcp

배포

Render

render.yaml 블루프린트가 포함되어 있습니다. 저장소에서 직접 배포하세요.

PM2

pm2 start ecosystem.config.cjs

API 엔드포인트

메서드	경로	설명
GET	`/health`	상태 확인
GET	`/mcp`	SSE 연결(새 세션 생성)
POST	`/mcp`	JSON-RPC 요청
GET	`/`	서비스 정보

클라이언트 구성

Claude Desktop

{
  "mcpServers": {
    "agent-browser": {
      "command": "node",
      "args": ["dist/index.js"]
    }
  }
}

원격 HTTP 클라이언트

{
  "mcpServers": {
    "agent-browser": {
      "url": "https://your-domain.com/mcp"
    }
  }
}

라이선스

MIT

Available Tools

39 tools

browser_a11yD

Accessibility audit

Name	Required	Description
`url`	No	URL to audit
`json`	No	JSON output
`tags`	No	axe tags filter
`session`	No	Session name
`selector`	No	Scope to element

browser_batchC

Execute multiple commands

Name	Required	Description	Default
`bail`	No	Stop on first error
`commands`	Yes	Array of commands to execute

browser_chatD

Natural language browser control

Name	Required	Description	Default
`session`	No	Session name
`instruction`	Yes	Natural language instruction

browser_clickC

Click element by selector or @ref

Name	Required	Description	Default
`session`	No	Session name
`selector`	Yes	CSS selector or @ref from snapshot

browser_closeC

Close browser

Name	Required	Description	Default
`all`	No	Close all sessions
`session`	No	Session name

browser_consoleC

View/clear console logs and errors

Name	Required	Description
`json`	No	JSON output
`action`	Yes	Console action
`session`	No	Session name

browser_cookieD

Manage cookies

Name	Required	Description
`name`	No	Cookie name
`value`	No	Cookie value
`action`	Yes	Cookie action
`session`	No	Session name

browser_dialogD

Handle dialogs

Name	Required	Description
`text`	No	Prompt text for accept
`action`	Yes	Dialog action
`session`	No	Session name

browser_diffC

Diff snapshots/screenshots

Name	Required	Description
`json`	No	JSON output
`type`	Yes	Diff type
`target`	No	Target URL
`session`	No	Session name
`baseline`	No	Baseline file/URL
`selector`	No	Scope to element

browser_doctorD

Diagnose install

Name	Required	Description	Default
`fix`	No	Run destructive repairs
`offline`	No	Skip network probes

browser_evalD

Execute JavaScript

Name	Required	Description	Default
`script`	Yes	JavaScript to execute
`session`	No	Session name

browser_fillC

Clear and fill input

Name	Required	Description
`text`	Yes	Text to fill
`session`	No	Session name
`selector`	Yes	CSS selector or @ref

browser_findC

Find elements semantically

Name	Required	Description
`name`	No	Accessible name filter
`type`	Yes	Find method
`value`	Yes	Value to find
`action`	Yes	Action to perform
`session`	No	Session name

browser_frameD

Switch frame

Name	Required	Description	Default
`session`	No	Session name
`selector`	No	Frame selector (omit for main)

browser_getC

Get text/html/value/url/title

Name	Required	Description
`session`	No	Session name
`property`	Yes	Property to get
`selector`	No	CSS selector (required for text/html/value)

browser_hoverD

Hover element

Name	Required	Description	Default
`session`	No	Session name
`selector`	Yes	CSS selector or @ref

browser_installC

Install Chrome

Name	Required	Description	Default
`deps`	No	Install system deps

browser_navigateC

Back/forward/reload

Name	Required	Description	Default
`session`	No	Session name
`direction`	Yes	Navigation direction

browser_networkD

Network requests/routes

Name	Required	Description
`url`	No	URL pattern for route/unroute
`body`	No	Mock response body
`type`	No	Filter by resource type (xhr, fetch, script, etc)
`abort`	No	Block requests
`action`	Yes	Network action
`filter`	No	Filter requests by text
`method`	No	Filter by HTTP method (GET, POST, etc)
`status`	No	Filter by status (200, 2xx, 4xx, etc)
`content`	No	HAR content embedding
`session`	No	Session name
`requestId`	No	Request ID for detail view

browser_openC

Open URL in browser

Name	Required	Description	Default
`url`	Yes	URL to navigate to
`session`	No	Session name for isolation

browser_pressC

Press key (Enter, Tab, etc)

Name	Required	Description	Default
`key`	Yes	Key to press (Enter, Tab, Control+a, etc)
`session`	No	Session name

browser_profileD

Chrome profiles

Name	Required	Description	Default
`name`	No	Profile name
`action`	Yes	Profile action

browser_reactD

React devtools

Name	Required	Description
`json`	No	JSON output
`action`	Yes	React action
`fiberId`	No	Fiber ID for inspect
`session`	No	Session name

browser_readC

Read page content or fetch URL

Name	Required	Description
`url`	No	URL to fetch (omit for active tab)
`filter`	No	Filter content sections
`session`	No	Session name

browser_screenshotD

Take screenshot

Name	Required	Description
`full`	No	Full page screenshot
`path`	No	Output file path
`session`	No	Session name

browser_scrollC

Scroll page

Name	Required	Description
`pixels`	No	Pixels to scroll
`session`	No	Session name
`selector`	No	Element to scroll
`direction`	Yes	Scroll direction

browser_selectC

Select dropdown option

Name	Required	Description
`value`	Yes	Option value
`session`	No	Session name
`selector`	Yes	CSS selector or @ref

browser_sessionD

Session management

Name	Required	Description
`scope`	No	Session scope
`action`	Yes	Session action
`prefix`	No	Session prefix

browser_setC

Set viewport/device/geo/media

Name	Required	Description
`value`	Yes	Setting value
`session`	No	Session name
`setting`	Yes	Setting to change

browser_skillsD

List/get skills

Name	Required	Description
`all`	No	Get all skills
`name`	No	Skill name
`action`	Yes	Skills action

browser_snapshotC

Get accessibility tree with refs

Name	Required	Description	Default
`session`	No	Session name

browser_stateC

Save/load auth state

Name	Required	Description
`path`	No	File path for save/load
`action`	Yes	State action
`session`	No	Session name

browser_storageC

Manage localStorage/sessionStorage

Name	Required	Description
`key`	No	Storage key
`type`	Yes	Storage type
`value`	No	Storage value
`action`	Yes	Storage action
`session`	No	Session name

browser_streamD

WebSocket streaming

Name	Required	Description
`port`	No	WebSocket port
`action`	Yes	Stream action
`session`	No	Session name

browser_tabD

Manage tabs

Name	Required	Description
`url`	No	URL for new tab
`label`	No	Label for new tab
`action`	Yes	Tab action
`target`	No	Tab id or label
`session`	No	Session name

browser_traceC

Record trace

Name	Required	Description
`path`	No	Output file path
`action`	Yes	Trace action
`session`	No	Session name

browser_typeC

Type text into input

Name	Required	Description
`text`	Yes	Text to type
`session`	No	Session name
`selector`	Yes	CSS selector or @ref

browser_vitalsD

Web Vitals metrics

Name	Required	Description
`url`	No	URL to measure
`json`	No	JSON output
`session`	No	Session name

browser_waitC

Wait for element/text/URL

Name	Required	Description
`url`	No	Wait for URL pattern
`text`	No	Wait for text
`session`	No	Session name
`timeout`	No	Timeout in ms
`selector`	No	Wait for element

Tool Schema Changelog

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

39 tool updatesv1.0.0Aug 27, 2026
- First observedbrowser_a11y
- First observedbrowser_batch
- First observedbrowser_chat
- First observedbrowser_click
- First observedbrowser_close
- First observedbrowser_console
- First observedbrowser_cookie
- First observedbrowser_dialog
- First observedbrowser_diff
- First observedbrowser_doctor
- First observedbrowser_eval
- First observedbrowser_fill
- First observedbrowser_find
- First observedbrowser_frame
- First observedbrowser_get
- First observedbrowser_hover
- First observedbrowser_install
- First observedbrowser_navigate
- First observedbrowser_network
- First observedbrowser_open
- First observedbrowser_press
- First observedbrowser_profile
- First observedbrowser_react
- First observedbrowser_read
- First observedbrowser_screenshot
- First observedbrowser_scroll
- First observedbrowser_select
- First observedbrowser_session
- First observedbrowser_set
- First observedbrowser_skills
- First observedbrowser_snapshot
- First observedbrowser_state
- First observedbrowser_storage
- First observedbrowser_stream
- First observedbrowser_tab
- First observedbrowser_trace
- First observedbrowser_type
- First observedbrowser_vitals
- First observedbrowser_wait

Deploy Server

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

C2.4/5.0

Scored 2026-08-27 04:01 across 39 tools

Disambiguation5/5

Each tool targets a distinct browser operation (e.g., click, fill, hover, scroll) with no overlapping responsibilities. Even similar tools like get and read are differentiated by specific data returning or page fetching, and snapshot is distinct from screenshot.

Naming Consistency5/5

All tools follow a consistent 'browser_<verb>' pattern (e.g., open, click, fill, type, press), with no mixed conventions or vague names. The naming is predictable and clearly conveys each tool's action.

Tool Count2/5

With 39 tools, the count exceeds the typical range and feels heavy for most workflows. While the browser automation domain is broad, many tools are granular (e.g., browser_console, browser_trace), leading to potential selection overhead.

Completeness4/5

The tool set covers a comprehensive range of browser interactions: navigation, element actions, state management, debugging, performance monitoring, and network control. Minor gaps exist (e.g., file upload, form submission explicit handling), but most core workflows are well-supported.

Maintenance

ActivityStale

ResponsivenessNo recent community issues

Resources

Related MCP Connectors

BrowserHere
app.1ts.browserhere
Control a real browser from AI agents: navigate, click, type, screenshot, read the page.
Browser Automation Agent Orchestration
1
https://browserhere.1ts.app/mcp
Irmu — Web Data MCP
io.github.polgoide
Web scraping, browser automation, screenshots and AI data extraction for AI agents via MCP.
Browser Automation Web Scraping Agent Orchestration
1
https://app.irmu.com/mcp
Skyvern
io.github.Skyvern-AI
AI-powered browser automation — navigate, click, fill forms, and extract data from any website.
Browser Automation Web Scraping App Automation
223,358
https://api.skyvern.com/mcp/
browserbasehq-mcp-browserbase
ai.smithery
Provides cloud browser automation capabilities using Stagehand and Browserbase, enabling LLMs to i…
Browser Automation Web Scraping Agent Orchestration
23,397
https://server.smithery.ai/@browserbasehq/mcp-browserbase/mcp

Related MCP Servers

openmcp
openmcp-pro
A
license
Not graded
quality
D
maintenance
Enables AI agents to automate web tasks such as browsing, clicking, typing, and taking screenshots via the Model Context Protocol.
Updated a year ago (2025-11-13 00:16 UTC)
1
MIT
selenium-mcp
Browser Automation Web Scraping Testing & QA Tools
gaforov
A
license
A
quality
B
maintenance
Enables browser automation through the Model Context Protocol, allowing AI agents to control Chrome, Firefox, or Edge for tasks like navigation, clicking, typing, and screenshots.
Updated 2 months ago (2026-09-13 02:00 UTC)
41
92 npm
MIT
Playwright Automation MCP
Browser Automation Testing & QA Tools
devqa07
A
license
A
quality
D
maintenance
Enables AI-powered browser automation controlled through natural language, integrating Playwright with the Model Context Protocol to perform web interactions like navigation, form filling, and screenshots.
Updated a year ago (2025-08-23 08:05 UTC)
10
MIT
Browseagent MCPofficial
Browser Automation Autonomous Agents
browseagent
A
license
A
quality
D
maintenance
Enables AI agents to control web browsers through the Model Context Protocol, supporting navigation, clicking, typing, and screenshots.
Updated a year ago (2025-09-20 15:33 UTC)
12
5 npm
1
MIT