Skip to main content
Glama
MCP Web Server
by zhaiy

MCP Web Server

本项目是一个本地 MCP Server,提供免费网络访问能力,可接入 Claude Desktop / Claude Code 等 MCP 客户端。

功能概览

  • http_request: 通用 HTTP 请求

  • web_search: 可配置搜索引擎(DuckDuckGo / Bing / 百度)

  • extract_webpage_content: 网页正文提取(含标题、列表、表格、代码块、引用)

  • fetch_json: 获取并解析 JSON

  • web_search_and_extract: 搜索 + 并发提取一体化

  • batch_http_request: 批量并发请求

  • screenshot_webpage(可选): 网页截图并返回 Base64(需安装 Playwright)

Related MCP server: web-search-mcp

安装

uv pip install -e .

开发依赖:

uv pip install -e ".[dev]"

启动服务

uv run python run_server.py

MCP 配置

Claude Code

claude mcp add web-server -s local -- uv run --directory /path/to/your/project python run_server.py

Claude Desktop

请参考仓库中的示例配置文件:claude_desktop_config.example.json。

使用前将 --directory 路径改为你的本地项目路径。

运行测试

uv pip install -e ".[dev]"
pytest

环境变量

变量名

说明

默认值

MCP_LOG_LEVEL

日志级别

INFO

MCP_HTTP_PROXY

HTTP 代理

空

MCP_HTTPS_PROXY

HTTPS 代理

空

MCP_USER_AGENT

自定义 User-Agent

内置 UA

MCP_DEFAULT_TIMEOUT

默认请求超时(秒)

30

MCP_SEARCH_ENGINE

搜索引擎:duckduckgo/bing/baidu

duckduckgo

MCP_BING_DOMAIN

Bing 检索域名(可设 www.bing.com/cn.bing.com)

www.bing.com

MCP_SEARCH_API_KEY

可选搜索 API Key(预留扩展)

空

MCP_SEARCH_API_ENDPOINT

可选搜索 API Endpoint(预留扩展)

空

MCP_RATE_LIMIT_SEARCH

web_search 每分钟调用上限

5

MCP_RATE_LIMIT_HTTP

http_request 每分钟调用上限

30

MCP_RATE_LIMIT_EXTRACT

extract_webpage_content 每分钟调用上限

10

MCP_STDIN_INVALID_INPUT_POLICY

非法 stdin 行处理策略:ignore/warn/strict

warn

MCP_STDIN_INVALID_INPUT_POLICY 说明:

  • warn(默认):忽略非法 JSON-RPC 行并记录 warning

  • ignore:忽略非法 JSON-RPC 行且不记录 warning

  • strict:收到非法 JSON-RPC 行即快速失败(便于定位客户端写入问题)

MCP_SEARCH_ENGINE 说明:

  • duckduckgo:默认选项,无 API Key;在部分网络环境下可能不可用。

  • bing:默认使用 www.bing.com 页面检索;可通过 MCP_BING_DOMAIN=cn.bing.com 切换大陆入口。

  • baidu:使用百度网页检索;中国大陆可用性通常较好。

可选 API Key 模式说明:

  • 当前版本默认使用“网页检索”模式(无需 API Key)。

  • MCP_SEARCH_API_KEY / MCP_SEARCH_API_ENDPOINT 已预留,便于后续接入官方 API 或第三方聚合 API。

可选截图依赖

uv pip install -e ".[screenshot]"
playwright install chromium

可选 Web 配置界面

项目内提供了一个基于 Flask 的辅助配置界面(web_config.py / run_web_config.py),用于通过浏览器引导执行 claude mcp add:

uv pip install -e ".[web]"
uv run python run_web_config.py

打开 http://localhost:5000 即可使用。
如果你只通过命令行配置 MCP,可忽略这两个文件。

架构图

flowchart TD
    Client[MCP Client] --> Server[FastMCP Server]
    Server --> Tools[Tools Layer]
    Tools --> Search[web_search / web_search_and_extract]
    Tools --> Http[http_request / batch_http_request / fetch_json]
    Tools --> Extract[extract_webpage_content]
    Tools --> Shot[screenshot_webpage optional]
    Http --> AsyncClient[Shared httpx.AsyncClient]
    Extract --> AsyncClient
    Search --> DDGS[DuckDuckGo via asyncio.to_thread]
    Search --> Bing[Bing HTML parsing]
    Search --> Baidu[Baidu HTML parsing + redirect resolve]

版本记录

详见 CHANGELOG.md。

许可证

本项目采用 Apache License 2.0 许可证。

Copyright 2025 MCP Web Server Contributors

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

Available Tools

6 tools
batch_http_requestD

Inputs: urls (required), method, headers, timeout, +1 more. Returns: object

extract_webpage_contentD

Inputs: url (required), max_length, include_links. Returns: object

fetch_jsonD

Inputs: url (required), timeout. Returns: object

http_requestD

Inputs: url (required), method, headers, timeout, +1 more. Returns: object

web_search_and_extractD

Inputs: query (required), region, num_results, max_content_length. Returns: object

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    D1.7/5.0

    Scored across 6 tools

    Disambiguation2/5

    Several tools have overlapping purposes: http_request and fetch_json both perform HTTP requests, and web_search and web_search_and_extract overlap in functionality. The distinctions between generic request and search/extract operations are unclear without detailed descriptions.

    Naming Consistency2/5

    Tool names mix different patterns: some start with verbs (fetch_json, extract_webpage_content), while others are noun phrases (http_request, web_search) or compound verbs (web_search_and_extract). No consistent verb_noun convention is followed.

    Tool Count5/5

    Six tools is a reasonable number for a web server utility, covering HTTP requests, JSON handling, batch operations, search, and extraction without being excessive or minimal.

    Completeness4/5

    The set covers core web operations: making requests, fetching JSON, batch processing, searching, and extracting content. Minor gaps exist, such as no explicit tool for HTML parsing beyond extraction, but the core surface is largely complete.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      B
      maintenance
      MCP server that lets Claude Code fetch web content using real Chrome browsers. Renders JavaScript-heavy pages, handles bot mitigation, and runs up to 14 parallel browsers locally with zero API keys. Makes outbound HTTP requests only to URLs the user explicitly asks Claude to fetch.
      2
      14 npm
      91
      MIT
    • F
      license
      A
      quality
      C
      maintenance
      A local MCP server that gives Claude Code web search capabilities using DuckDuckGo, enabling discovery of URLs from natural language queries.
      1
      1
      -