About this project

zvec-grep (zg) is a local-first search layer for workspaces, designed for both human users and AI agents. It combines ripgrep, BM25, and vector search behind a single interface, allowing users to search by meaning, rank results by relevance, and verify findings with exact text or regex. The tool supports source code, documents, and structured data while preserving structure and source locations. Key features include CLI and agent integrations, multi-format search, ranked source-linked results, and local-by-default operation where files, indexes, and local models remain on the user's machine. Remote embeddings are used only with permission. The project provides a TypeScript/Node.js implementation, with a separate Rust implementation under development. Installation requires Node.js 22 or newer. Users can install globally via npm, index a workspace with `zg index`, and query it with `zg query`. Agent integrations are available for OpenCode, Codex, Claude Code, Qwen Code, Qoder, Cursor, GitHub Copilot, VS Code, Grok Build, and others. The tool also supports MCP for secure agent access. Documentation covers agent integrations, CLI usage, MCP, retrieval pipeline, architecture, server and execution modes, embedding models, and roadmap. Benchmarks include cross-domain agent evaluations on SWE-QA-Bench and BrowseComp-Plus, plus real-world case studies with Pylint, Matplotlib, and Django repositories.