mrkrsl/web-search-mcp

Health v2 · maintenance only

Adoption not part of the score

1109 stars · 161 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A locally hosted MCP (Model Context Protocol) server written in TypeScript that gives local LLMs web search capabilities without requiring API keys. It searches via Bing, Brave, and DuckDuckGo using a mix of browser automation (Playwright) and HTTP requests, and offers tools for full-page content extraction, snippet-only search, and single-page content retrieval.

Use cases

• give my local llm web search without api keys
• mcp server for lm studio web search
• extract full webpage content for rag with local models
• search the web from librechat with a local model
• self-hosted web search tool for mcp clients
• scrape search results and page content for an ai agent

When to choose

• you run local LLMs in LM Studio or LibreChat and want tool-based web search
• you want to avoid paid search APIs and API keys
• you need full-page content extraction, not just snippets
• you prefer a self-hosted, MIT-licensed solution

When to avoid

• you need a cloud-hosted or managed search service
• your MCP client is untested with this server and you need guaranteed compatibility
• you rely on older Llama or DeepSeek R1 models that handle tool calls poorly
• you need high-volume scraping at scale with proxy rotation

Facets

service · maturity active

mcp web-scraping search-engine llm-inference agent-framework large-language-models self-hosted developer-tools self-hosted cross-platform mcp-server local-llm web-search lm-studio no-api-keys playwright bing duckduckgo brave-search search ai-agents nodejs

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mrkrsl/web-search-mcp")