Codesearch

关于

Codesearch provides a powerful, fully offline solution for semantic code search, allowing developers to query their codebases using natural language. Built with Rust for speed, it leverages advanced techniques like Tree-sitter AST-aware chunking and ONNX models for local processing, ensuring privacy and sub-second search times. It supports incremental indexing for rapid updates and offers flexible indexing options (local or global). Crucially, codesearch acts as an efficient MCP server, offering token-optimized tools for AI agents like OpenCode and Claude Code, facilitating a "search → narrow → read" workflow for enhanced productivity.

主要功能

• Incremental Indexing for fast codebase updates
• MCP Server for AI Agent Integration (OpenCode/Claude Code)
• Tree-sitter AST-aware Smart Chunking
• Semantic & Hybrid Search with Neural Reranking
• 0 GitHub stars
• Fully Local, Offline & Private processing via ONNX models

使用案例

• Efficiently navigate large codebases and understand their structure without relying on external APIs.
• Integrate with AI coding assistants to provide contextual, token-efficient code search and reference finding.
• Quickly find relevant code sections using natural language queries (e.g., "where do we handle authentication?").