haihengh/llm-search repository overview

⁠LLM Search

Give your local LLM internet search capability — no API keys, no rate limits, no signups.

⁠Features

• web_search — searches via self-hosted SearXNG (Google, Bing, DDG, etc.)
• fetch_page — reads full web page content from URLs
• Streaming SSE — token-by-token output
• Anthropic API — /v1/messages for native Claude Code integration
• OpenAI API — /v1/chat/completions for any OpenAI-compatible client
• MCP Server — Model Context Protocol for Claude Desktop
• LM Studio + Ollama — supports both local LLM backends

⁠Quick Start

git clone https://github.com/haihengh/llm-search
cd llm-search && docker compose up -d

⁠Claude Code

export ANTHROPIC_BASE_URL=http://localhost:8000
export ANTHROPIC_AUTH_TOKEN=not-needed
export CLAUDE_CODE_ATTRIBUTION_HEADER=0
claude "What is the latest Go version?"

GitHub: https://github.com/haihengh/llm-search⁠