BetterWebSearch MCP

BetterWebSearch MCP — smarter web search, better data extraction, flexible and extensible, open source

Keyless-first. Works out of the box with DuckDuckGo — no API keys. Add BRAVE_API_KEY later for richer ranking & recency.

BetterWebSearch MCP is a local Model Context Protocol server that gives AI agents genuinely better web research: multi-provider web search, three-tier page extraction, and deep research with citations — without locking you into a single search API and without an API key to get started.

— npm downloads — last 7 days — GitHub stars 0.10.0 latest version

Quick start (30 seconds)#

Add this to your MCP client config (Claude Desktop, Claude Code, Cursor, VS Code Copilot, OpenCode, or any MCP client):

{
  "mcpServers": {
    "better-web-search-mcp": {
      "command": "npx",
      "args": ["-y", "better-web-search-mcp"]
    }
  }
}

That’s it. Every search tool works immediately. No .env, no API keys, no setup.

What makes it better#

  • Search Better — Aggregated search across DuckDuckGo, Brave, and Tavily — deduplicated, reranked, and query-expanded for better coverage.
  • Extract Smarter — Three-tier extraction pipeline: fast HTTP, structured data hydration, then full Playwright browser render — only escalates when needed.
  • Deep Research — Rewrites your question into multiple variants, searches in parallel, extracts top results with citations and confidence scores.
  • Cheaper Context — Doing the reading inside the server rather than pasting pages into the model cuts payload by 86.5% on the shipped benchmark, while keeping 97.1% of the answers.

The six tools#

ToolWhat it does
web_searchAggregated search across providers, deduplicated and reranked
web_researchDeep research: rewrites the question, searches in parallel, extracts with citations
deep_searchAlias for web_research
web_extractClean content from any URL, using the cheapest method that works
web_findSearch scoped to a single website
web_newsRecent news with a timeline and domain diversity

Works with#

Any MCP client speaking stdio — including Claude Desktop, Claude Code, Cursor, VS Code Copilot, Windsurf, Cline, Zed and OpenCode. See the quickstart for per-client config snippets.

Frequently asked#

No. DuckDuckGo is the default provider and needs no key, no account and no billing. BRAVE_API_KEY or TAVILY_API_KEY are optional and add publication dates and stronger ranking.

Is it free?#

Yes — MIT licensed, and the default provider costs nothing to query.

How is it different from other web search MCP servers?#

Most wrap a single search API and hand raw results back. This one aggregates providers, expands the query into variants, escalates extraction only as far as a page demands, clusters near-duplicate sources so a syndicated wire story is not mistaken for five independent confirmations, and returns cited passages instead of whole pages.

Does it run in the cloud?#

No. It runs locally over stdio, next to your client. Nothing is proxied through a third-party service.

Documentation#

Edit Edit this page