Codeweave

This server gives your AI agent structured code understanding instead of dumping entire files into context. It runs locally, parses your codebase with tree-sitter across seven languages, and exposes three tools: semantic_search for hybrid vector and keyword queries with density-based reranking, reindex for manual updates, and get_index_status for health checks. The search pipeline combines Qwen3 embeddings via Ollama with BM25 full-text search, then reranks results using call graph centrality, visibility, and structural complexity signals to surface important code over boilerplate. Setup runs through npx with an interactive wizard that handles Ollama installation and project indexing. Reach for this when you want cheaper, more relevant context for coding tasks without blowing through tokens on file dumps.