Zilliz drops Claude Context for semantic MCP search

This tool signals the end of brute-force context windows, shifting AI coding toward surgical, AST-aware retrieval to maintain model reasoning quality.

• –Incremental indexing via Merkle trees ensures only modified files are re-processed, slashing API costs and processing time.
• –Tree-sitter integration ensures code snippets remain structurally valid and logically coherent after chunking.
• –Reported 40% token savings by feeding models only the most relevant context for a given task.
• –Open-source stack supports Milvus, Zilliz Cloud, and major embedding providers like OpenAI and VoyageAI.