Semble: AI agent code search tool reduces token usage by 98%
Original: Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep
Why This Matters
Addresses critical efficiency bottleneck in AI agent code analysis workflows
MinishLab released Semble, a code search library for AI agents that uses 98% fewer tokens than grep+read operations. The tool indexes and searches entire codebases in under one second with 200x faster indexing than code-specialized transformers.
Semble is an open-source code search library specifically designed for AI agents, developed by MinishLab. The tool dramatically reduces token consumption by approximately 98% compared to traditional grep+read operations while maintaining 99% retrieval quality. It features end-to-end codebase indexing and searching in under one second, with 200x faster indexing speeds and 10x faster queries than code-specialized transformers. The library is available on GitHub with 1.5k stars and includes multiple interfaces including MCP Server, CLI, Python API, and bash integration. Semble aims to improve agent efficiency by providing exact code snippets instantly with significantly reduced latency.