Semble is a fast, CPU-only code search library designed for AI coding agents. It combines static Model2Vec embeddings with BM25 lexical search and Reciprocal Rank Fusion to deliver semantic code search that uses ~98% fewer tokens than grep+read. Indexing a full repo takes ~250ms and queries resolve in ~1.5ms, achieving NDCG@10 of 0.854 — 99% of the quality of the much larger CodeRankEmbed Hybrid model. It integrates as an MCP server for Claude Code, Cursor, Codex, and OpenCode, or via bash/AGENTS.md. A Python API is also available for programmatic use. No GPU, API keys, or external services required.

9m read timeFrom github.com
Post cover image
Table of contents
Fast and Accurate Code Search for Agents Uses ~98% fewer tokens than grep+readQuickstartMain FeaturesMCP ServerBash integrationCLIPython APIHow it worksBenchmarksLicenseCiting

Sort: