New Tool Semble Offers Efficient Code Search with Reduced Token Usage

May 17, 2026

AI Summary

Stephan and Thomas have launched Semble, an open-source code search tool that significantly reduces token usage compared to traditional methods like grep. It achieves high retrieval quality and speed, making it a viable solution for searching large codebases without the need for external services or configurations.

Semble is an open-source code search tool developed by Stephan and Thomas.
It addresses inefficiencies in existing code search methods, particularly when fallback to grep is required.
The tool uses Model2Vec embeddings and BM25 for improved search efficiency, operating entirely on CPU without transformers.
In benchmarks involving 1250 query/document pairs across 63 repositories and 19 programming languages, Semble demonstrated 98% fewer token usage compared to grep and maintained 99% of the retrieval quality of a larger transformer model.
Key features include fast indexing (approximately 250ms for typical repositories) and quick query responses (around 1.5ms per query).
Semble can be integrated as a drop-in replacement for various coding tools and requires no API keys or GPU.
Users can install Semble in Claude Code with a simple command and access additional information through its README and benchmarks on GitHub.

code searchopen sourceefficiencyagentssoftware development

New Tool Semble Offers Efficient Code Search with Reduced Token Usage

Related Stories

Google announces new AI-powered audio glasses in partnership with Warby Parker and Gentle Monster

Google introduces new AI models and agents to enhance user experience and compete in the market

Google introduces customizable AI agents for continuous information updates

Google I/O 2026 Introduces New AI Models and Enhanced Agent Capabilities