OneSearch MCP Server
The OneSearch MCP Server is a versatile tool for web search, scraping, crawling, and structured data extraction, integrating multiple search providers and browser automation.
Web Search (
one_search): Search using SearXNG, Tavily, DuckDuckGo, Bing, Google, Zhipu, Exa, Bocha, or local browser automation (Bing, Google, Baidu, Sogou). Supports filtering by language, time range, and categories (news, images, videos, science, etc.).Web Scraping (
one_scrape): Scrape single pages with output in markdown, HTML, raw HTML, screenshots, or links. Supports browser actions (click, scroll, type, execute JavaScript), mobile viewport emulation, and geolocation settings.URL Discovery (
one_map): Discover URLs from a starting point via sitemap.xml and/or HTML link crawling, with options to filter by search term, include subdomains, and limit results.Structured Data Extraction (
one_extract): Extract structured information from multiple URLs simultaneously using LLM-powered extraction with custom prompts, JSON schemas, and optional web search context.Local/Free Operation: Run browser-automated searches and scraping without API keys using
agent-browser(requires a Chromium-based browser), ensuring privacy and no API costs.Multi-provider Flexibility: Easily switch between free (DuckDuckGo, local) and premium (Tavily, Exa, Bocha, Zhipu) providers via environment variables. Available via npm, Docker, and compatible with Claude Desktop, Cursor, and Windsurf.
Used for installation and distribution of the MCP server package, with support for running the server through npx commands.
Enables web search capabilities through SearXNG, allowing queries to be performed across multiple search engines through a single interface.
Supports installation and execution through shell commands, with environment variable configuration for different search providers and APIs.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@OneSearch MCP Server find recent articles about quantum computing breakthroughs"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
π OneSearch MCP Server: Web Search, URL Discovery, Scraper & Content Prep
A Model Context Protocol (MCP) server implementation that integrates with multiple search providers for web search, local browser search, URL discovery, and scraping capabilities with agent-browser.
Features
Web search, scrape, discover URLs, and preprocess content from websites.
Support multiple search engines and web scrapers: SearXNG, Tavily, DuckDuckGo, Bing, Google, Zhipu (ζΊθ°±), Exa, Bocha (εζ₯), You.com, etc.
Local web search (browser search), support multiple search engines: Bing, Google, Baidu, Sogou, etc.
Use
agent-browserfor browser automation.Free, no API keys required.
Enabled tools:
one_search,one_scrape,one_map,one_extract
Related MCP server: Firecrawl MCP Server
Current Tool Surface
one_searchReturns search results from the configured provider.
one_mapDiscovers links from a starting URL by loading the page in the browser and extracting links from its HTML.
Supported input fields:
url,search,includeSubdomains,limit.This is not a sitemap crawler; removed fields such as
ignoreSitemapandsitemapOnlyare rejected at the schema boundary.
one_scrapeScrapes one page and returns content selected by
formats.Supported input fields:
url,formats,waitFor,timeout,skipTlsVerification,allowExecuteJavascript,actions.Supported formats:
markdown,html,rawHtml,links,screenshot,screenshot@fullPage.Supported bounded pre-scrape actions:
wait,click,write,press,scroll.Advanced pre-scrape action:
executeJavascript. IfactionscontainsexecuteJavascript, you must setallowExecuteJavascript: true.actionsrun serially before content capture and fail fast on the first action error.Removed fields such as
onlyMainContent,extract, andlocationare rejected at the schema boundary.
one_extractAccepts only
urlsand returns preprocessed text blocks for downstream tools or models.
Migration from v1.1.0 and Earlier
Breaking Changes in v1.1.0:
Firecrawl Removed: The Firecrawl integration has been removed in favor of
agent-browser, which provides similar functionality without requiring external API services.New Browser Requirement: You must install Chromium browser (see Prerequisites section).
Environment Variables:
FIRECRAWL_API_URLandFIRECRAWL_API_KEYare no longer used.
What Changed:
one_scrapeandone_mapnow useagent-browserinstead of Firecrawlone_extractnow preprocesses multi-URL page content for downstream analysis instead of performing built-in LLM extractionAll browser-based operations are now handled locally, providing better privacy and no API costs
Migration Steps:
Install Chromium browser (see Prerequisites)
Remove
FIRECRAWL_API_URLandFIRECRAWL_API_KEYfrom your environment variablesUpdate to the latest version:
npm install -g one-search-mcp@latest
Prerequisites
Browser Requirement: This server uses agent-browser for web scraping and local search, which requires a Chromium-based browser.
Good News: The server will automatically detect and use browsers already installed on your system:
β Google Chrome
β Microsoft Edge
β Chromium
β Google Chrome Canary
If you don't have any of these browsers installed, you can:
# Option 1: Install Google Chrome (Recommended)
# Download from: https://www.google.com/chrome/
# Option 2: Install Microsoft Edge
# Download from: https://www.microsoft.com/edge
# Option 3: Install Chromium via agent-browser
npx agent-browser install
# Option 4: Install Chromium directly
# Download from: https://www.chromium.org/getting-involved/download-chromium/Installation
Using Claude Code CLI (Recommended)
# Add to Claude Code with default settings (local search)
claude mcp add one-search-mcp -- npx -y one-search-mcp
# Add with custom search provider (e.g., SearXNG)
claude mcp add one-search-mcp -e SEARCH_PROVIDER=searxng -e SEARCH_API_URL=http://127.0.0.1:8080 -- npx -y one-search-mcp
# Add with Tavily API
claude mcp add one-search-mcp -e SEARCH_PROVIDER=tavily -e SEARCH_API_KEY=your_api_key -- npx -y one-search-mcpManual Installation
# Install globally (Optional)
npm install -g one-search-mcp
# Or run directly with npx
npx -y one-search-mcpLocal Debugging with MCP Inspector
Use the official MCP Inspector to interactively test tools, resources, and prompts against this server during local development.
# Inspect the TypeScript source entrypoint
npm run inspector
# Inspect the built server entrypoint
npm run inspector:build
# Example with explicit environment variables
ALLOW_PRIVATE_NETWORK=true SEARCH_PROVIDER=local npm run inspectorUsing Docker
Docker image includes all dependencies (Chromium browser) pre-installed, no additional setup required.
Pull the image:
# From GitHub Container Registry
docker pull ghcr.io/yokingma/one-search-mcp:latest
# Or from Docker Hub
docker pull zacma/one-search-mcp:latestConfigure with Claude Desktop:
{
"mcpServers": {
"one-search-mcp": {
"command": "docker",
"args": ["run", "-i", "--rm", "ghcr.io/yokingma/one-search-mcp:latest"],
"env": {
"SEARCH_PROVIDER": "local"
}
}
}
}With custom search provider:
{
"mcpServers": {
"one-search-mcp": {
"command": "docker",
"args": [
"run", "-i", "--rm",
"-e", "SEARCH_PROVIDER=tavily",
"-e", "SEARCH_API_KEY=your_api_key",
"ghcr.io/yokingma/one-search-mcp:latest"
]
}
}
}Environment Variables
Search Provider
SEARCH_PROVIDER (Optional): The search provider to use, supports
searxng,duckduckgo,bing,tavily,google,zhipu,exa,bocha,ydc,local, default islocal.SEARCH_API_URL (Optional): The URL of the SearxNG API, or Google Custom Search Engine ID for
google.SEARCH_API_KEY (Optional): The API key for the search provider, required for
tavily,bing,google,zhipu,exa,bocha.YDC_API_KEY (Optional): You.com API key for
SEARCH_PROVIDER=ydc.
// supported search providers
export type SearchProvider = 'searxng' | 'duckduckgo' | 'bing' | 'tavily' | 'google' | 'zhipu' | 'exa' | 'bocha' | 'ydc' | 'local';Browser Network Security
ALLOW_PRIVATE_NETWORK (Optional): Set to
trueto allow browser-backed tools (one_scrape,one_map,one_extract) to access private, loopback, and link-local network targets. Defaults tofalse.
Use ALLOW_PRIVATE_NETWORK=true only in trusted deployments. When enabled, prompt injection or untrusted tool inputs can make the MCP process fetch internal services that are otherwise blocked by default.
Search Provider Configuration
Provider | API Key Required | API URL Required | Notes |
| No | No | Free, uses browser automation |
| No | No | Free, no API key needed |
| Optional | Yes | Self-hosted meta search engine |
| Yes | No | |
| Yes | No | |
| Yes | Yes (Search Engine ID) | |
| Yes | No | |
| Yes | No | |
| Yes | No | |
| Yes | No |
Configuration for Other MCP Clients
Claude Desktop
Add to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"one-search-mcp": {
"command": "npx",
"args": ["-y", "one-search-mcp"],
"env": {
"SEARCH_PROVIDER": "local"
}
}
}
}Cursor
Add to your mcp.json file:
{
"mcpServers": {
"one-search-mcp": {
"command": "npx",
"args": ["-y", "one-search-mcp"],
"env": {
"SEARCH_PROVIDER": "local"
}
}
}
}Windsurf
Add to your ./codeium/windsurf/model_config.json file:
{
"mcpServers": {
"one-search-mcp": {
"command": "npx",
"args": ["-y", "one-search-mcp"],
"env": {
"SEARCH_PROVIDER": "local"
}
}
}
}Self-hosting SearXNG (Optional)
If you want to use SearXNG as your search provider, you can deploy it locally using Docker:
Prerequisites:
Docker installed and running (version 20.10.0 or higher)
At least 4GB of RAM available
Quick Start:
# Clone SearXNG Docker repository
git clone https://github.com/searxng/searxng-docker.git
cd searxng-docker
# Start SearXNG
docker compose up -dAfter deployment, SearXNG will be available at http://127.0.0.1:8080 by default.
Configure OneSearch to use SearXNG:
# Set environment variables
export SEARCH_PROVIDER=searxng
export SEARCH_API_URL=http://127.0.0.1:8080For more details, see the official SearXNG Docker documentation.
Troubleshooting
Browser not found error
If you see an error like "Browser not found", the server couldn't detect any installed Chromium-based browser. Please install one of the following:
Google Chrome: https://www.google.com/chrome/
Microsoft Edge: https://www.microsoft.com/edge
Chromium: https://www.chromium.org/getting-involved/download-chromium/
Or install via agent-browser:
npx agent-browser installLicense
MIT License - see LICENSE file for details.
Available Tools
4 toolsone_extractA
Fetch and preprocess page content from one or more URLs. Returns cleaned text blocks that can be passed to downstream tools or models.
Inputs: urls (required).
one_mapA
Discover URLs from a starting point by loading a page in the browser and extracting links from its HTML.
Inputs: url (required), limit, search, includeSubdomains.
one_scrapeA
Scrape a single webpage and return markdown, HTML, links, or a screenshot. Supports navigation timeout, TLS verification control, full-page screenshots, bounded pre-scrape actions, and advanced executeJavascript only when allowExecuteJavascript is true.
Inputs: url (required), actions, formats, timeout, +3 more.
one_searchB
Search and retrieve content from web pages. Returns SERP results by default (url, title, description).
Inputs: query (required), limit, language, timeRange, +1 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelogβ¦
TDQS
Scored across 4 tools
Each tool has a distinct core function: search returns SERP results, extract preprocesses multiple URLs, scrape handles single-page advanced operations, and map discovers links. There is minor overlap between extract and scrape for fetching page content, but their descriptions make the intended use cases clear.
All tool names follow the consistent pattern 'one_' prefix followed by a lowercase action verb (search, extract, scrape, map), with underscores between words. No mixed conventions or camelCase.
Four tools is well within the ideal range for a focused search/scraping server. Each tool covers a distinct aspect of web research, making the set feel appropriately scoped without being bloated.
The tool set covers the primary workflows: searching, extracting content, single-page scraping with advanced options, and link discovery. Minor gaps exist, such as no dedicated batch scraping tool, but the combined functionality handles most common use cases.
Maintenance
Related MCP Connectors
Web search API and MCP server for developers building AI agents.
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to interact with yoβ¦
Free web search for AI agents. No API key required. Hosted MCP in active development.
One MCP server for all web data. Scrape, search, crawl, and AI research in one install.
Related MCP Servers
- FlicenseCqualityCmaintenanceBuilt as a Model Context Protocol (MCP) server that provides advanced web search, content extraction, web crawling, and scraping capabilities using the Firecrawl API.41-
- AlicenseBqualityDmaintenanceA Model Context Protocol server that enables AI assistants to perform advanced web scraping, crawling, searching, and data extraction through the Firecrawl API.9MIT
- AlicenseAqualityBmaintenanceA Model Context Protocol server that provides real-time web search capabilities to AI assistants through pluggable search providers, currently integrated with the Brave Search API.534 npm16MIT
- AlicenseAqualityDmaintenanceA Model Context Protocol server that enables web scraping, crawling, and content extraction capabilities through integration with Firecrawl.82MIT