This project bridges LLMs with live web search by acting as a proxy layer. It intercepts LLM requests and injects search results when the model determines browsing is needed. Supports multiple LLM providers (OpenAI, Gemini, Moonshot, Groq) and search engines (Google, Bing, DuckDuckGo, SearXNG). The model decides autonomously when to trigger web lookups without requiring plugins, using function calling to orchestrate the search-to-response workflow.