mcp-browser-search
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| BROWSER_CDP_PORT | No | CDP debugging port | 9222 |
| BROWSER_SEARCH_EXE | No | Browser executable file path (auto-detected by default) | |
| BROWSER_SEARCH_ENGINE | No | Search engine to use: google / bing / baidu | |
| BROWSER_SEARCH_PROFILE | No | Dedicated browser profile directory; login state is stored here | ~/.mcp-browser-profile |
| BROWSER_SEARCH_AUTOSTART | No | Set to 0 to disable automatically launching the browser | 1 |
| BROWSER_SEARCH_ALLOW_PRIVATE | No | Set to 1 to allow access to intranet / localhost addresses | 0 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| web_searchA | 联网搜索(走用户已登录的真实浏览器,百度/Bing)。返回标题、URL、摘要。用于查资料、找文档、确认事实。 |
| web_openA | 用浏览器打开一个网址,返回页面标题和正文纯文本。走用户已登录的浏览器,所以能看到登录后才有内容。 |
| web_readA | 读取当前浏览器标签页的正文(不导航)。适合点完链接或翻页之后取内容。 |
| web_clickA | 在当前页面点击一个元素(CSS 选择器),点完等页面稳定并返回新页面信息。 |
| web_typeA | 在当前页面的输入框里填内容(CSS 选择器),可选是否提交(优先真正提交所在表单,没有表单才发真回车键)。 |
| browser_statusB | 检查浏览器通道是否可用:CDP 端口、浏览器是否在跑、进程内配置。 |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 6 tools
Each tool maps to a distinct action: web_search finds pages, web_open navigates to a URL, web_read extracts the current tab without navigating, and web_click/web_type operate on specific elements. The read-vs-open distinction is explicitly clarified, though a naive agent might still conflate web_search with web_open when simply wanting page content.
Five tools follow a clean web_<verb> pattern (web_search, web_read, web_open, web_click, web_type). browser_status deviates from the prefix but remains self-explanatory and semantically distinct as a diagnostics tool.
Six tools is a well-scoped surface for a browser automation server—each tool (search, open, read, click, type, status check) earns its place in the interaction loop with no redundancy.
The core loop of search → open → read → interact (click/type) is fully covered, plus a channel health check. Minor gaps remain: no back/scroll navigation, no tab management or screenshot, and no explicit key-press or select-element handling, but agents can work around these.