n24q02m/wet-mcp
🔌 MCP Servern24q02m
A self-hosted MCP server for web search, content extraction, and hybrid document indexing without API keys.
wet-mcp serves as a comprehensive bridge between AI agents and the web, offering a suite of tools designed for information retrieval and knowledge management. At its core, the server integrates an embedded SearXNG instance, allowing for privacy-focused web searches directly from an MCP-compatible client. Beyond simple search, it features advanced content extraction capabilities, enabling agents to parse and process web pages effectively.
One of its standout technical features is the hybrid search engine, which combines FTS5 (Full-Text Search) with semantic search. This dual approach ensures that agents can retrieve information based on both exact keyword matches and conceptual relevance. The inclusion of the Qwen3 embedding model directly within the server architecture allows for local vectorization, removing dependencies on third-party embedding APIs. This architecture is particularly beneficial for developers building local-first AI applications, as it ensures data sovereignty and reduces latency. The server is highly modular, allowing for seamless integration into existing MCP workflows, and its design emphasizes low-overhead operation, making it suitable for both local development environments and self-hosted production setups.
💡Highlights
- ├─Embedded SearXNG search engine
- ├─Hybrid FTS5 and semantic search
- └─Local Qwen3 embedding, no API keys
🎯For
- ├─AI Developers
- └─RAG Engineers