SearXNG AI Kit Review 2026

A standalone CLI, Python library, and MCP server that packages the SearXNG privacy-respecting metasearch engine — 180+ search engines with AI research features, no server setup needed.

SearXNG AI Kit takes SearXNG — the popular privacy-respecting metasearch engine that normally requires running your own server — and packages it as a standalone command-line tool, Python library, and MCP server. There is nothing to host: download a pre-built binary for Linux or macOS (or install from source with uv), and you can immediately search across 180+ engines including DuckDuckGo, Startpage, Brave, Google, and Bing, with categories for general web, images, videos, news, science, IT, music, and files. MCP support is a headline feature: running `searxng mcp-server` exposes five tools to AI assistants like Claude — web_search, multi_web_search (parallel queries), fetch_url and fetch_urls (clean content extraction via Jina.ai), and an `ask` tool that gives a model recursive access to all of the above — with a ready-made Claude Desktop config snippet in the README. Beyond MCP, the kit has its own AI layer: `searxng ask` for one-shot research questions and `searxng chat` for interactive sessions with conversation memory and saved markdown history, backed by a model registry supporting OpenAI, Anthropic, Google, OpenRouter, and — via CLI Proxy API integration — your existing Claude Pro, ChatGPT Plus, or Gemini subscriptions with no API keys. The Python library mirrors everything programmatically, with sync and async variants. Output is available as human-readable text or JSON for scripting. Free and open source; SearXNG itself is AGPL-3.0 licensed.