searxng-mcp
MCP server that connects to a self-hosted SearXNG instance and exposes web search capabilities over MCP HTTP (JSON-RPC), enabling AI chatbots to search the internet for real-time information.
README
searxng-mcp
MCP server that connects to a self-hosted SearXNG instance and exposes web search capabilities over MCP HTTP (JSON-RPC).
Overview
This service enables your AI chatbot to search the internet for real-time information via SearXNG, a privacy-respecting metasearch engine.
Use cases:
- Current weather, news, prices
- Recent documentation and updates
- Real-time information beyond the LLM's knowledge cutoff
Environment
Required:
SEARXNG_URL- Base URL of your SearXNG instance (e.g.http://searxng:8080).MCP_INTERNAL_TOKEN- Shared secret betweenhomelab-chatbot-uiandsearxng-mcp. Alias:SEARXNG_MCP_INTERNAL_TOKEN.
Optional:
PORT- Server port. Defaults to3003.SEARXNG_TIMEOUT_MS- Request timeout in milliseconds. Defaults to10000.
Tools
web_search
Search the internet for current information.
Parameters:
| Name | Type | Required | Description |
|---|---|---|---|
query |
string | Yes | The search query. Be specific for better results. |
category |
enum | No | Search category: general, news, images, videos, science, it, files, music. Default: general. |
engines |
string[] | No | Specific search engines (e.g. ["google", "duckduckgo", "brave"]). |
language |
string | No | Language code for results (e.g. en, zh, ja). |
limit |
number | No | Maximum results (1-20). Default: 5. |
Response:
{
"status": "ok",
"query": "weather hong kong",
"resultCount": 5,
"results": [
{
"title": "Hong Kong Weather Forecast",
"url": "https://example.com/weather",
"snippet": "Current conditions: 24°C, partly cloudy...",
"source": "google",
"date": "2024-01-15"
}
],
"formatted": "[1] Hong Kong Weather Forecast\n URL: https://example.com/weather\n Current conditions: 24°C..."
}
Endpoints
| Method | Path | Description |
|---|---|---|
GET |
/health |
Health check (no auth required) |
GET |
/mcp |
SSE endpoint for streaming MCP clients |
POST |
/mcp |
Streamable HTTP endpoint for stateless MCP requests |
POST |
/ |
Legacy endpoint for backward compatibility |
Authentication
All MCP endpoints require authentication via:
Authorizationheader:Bearer <token>or plain<token>x-mcp-internal-tokenheader: plain<token>
Development
# Install dependencies
npm install
# Build
npm run build
# Run (requires env vars)
SEARXNG_URL=http://localhost:8080 \
SEARXNG_MCP_INTERNAL_TOKEN=test-token \
npm start
# Run tests
npm test
Docker
# Build image
docker build -t searxng-mcp .
# Run container
docker run -p 3003:3003 \
-e SEARXNG_URL=http://searxng:8080 \
-e SEARXNG_MCP_INTERNAL_TOKEN=your-secret-token \
searxng-mcp
Kubernetes
Deploy alongside your SearXNG instance:
# ConfigMap
apiVersion: v1
kind: ConfigMap
metadata:
name: searxng-mcp-config
data:
PORT: "3003"
SEARXNG_URL: "http://searxng.app-searxng.svc.cluster.local:8080"
SEARXNG_TIMEOUT_MS: "10000"
See deploy/k8s/base/ for full manifests.
Registering in Chatbot UI
- Deploy this service
- Go to Admin → Tools → Add Backend
- Configure:
- Type:
mcp - Transport:
http - URL:
http://searxng-mcp.mcp-searxng.svc.cluster.local:3003/mcp - Auth Type:
apiKey - API Key: Your
SEARXNG_MCP_INTERNAL_TOKENvalue
- Type:
- Click "Discover Tools" to auto-import the
web_searchtool - Enable the tool
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。