Show HN: Semble – Fast code search for agents with near-transformer accuracy
Category: library
Tags: code-search, agent-tools, mcp-server, embeddings, bm25, python
Score: 7.8/10 (Innovation: 7, Technical: 8, Documentation: 8, Utility: 8)
Semble is a fast code search library for AI agents that indexes and searches codebases in milliseconds on CPU with near-transformer accuracy. It uniquely combines static embeddings with BM25 and code-aware reranking, achieving 99% of the retrieval quality of transformer-based CodeRankEmbed while being 200x faster to index and 10x faster to query. Its MCP server integration makes it a practical tool for agents like Claude Code and Cursor to perform instant, offline code search without external APIs.
Target audience: backend devs, devops, data engineers
Repository: https://github.com/MinishLab/semble · Python · MIT · 4 stars
View on Hacker News