zvec-grep (zg)

Screenshot

About zvec-grep (zg)

zvec-grep (zg), powered by the zvec vector database, combines ripgrep, BM25 ranked search, and vector (semantic) search behind a single local-first interface. Use it directly from the terminal or let your AI agent use it. Discover files by meaning, rank results by relevance, then verify with exact text or regex. It searches source code, documents, and structured data while preserving source locations, so results are ranked, source-linked, and compact, which means fewer tool calls, fewer tokens, and less noise for agents. Code is indexed with symbols, signatures, and breadcrumbs, while prose is split into focused sections. It integrates with Codex, Claude Code, Qwen Code, Qoder, Cursor, GitHub Copilot, VS Code, and OpenCode, and exposes its tools through MCP. Files, indexes, and local embedding models stay on your machine by default, and remote embeddings receive data only with your permission. It supports local or remote embedding models, runs as one-off commands or a local server, works on macOS, Linux, and Windows (Node.js 22+), and is free and open source under the Apache 2.0 License.

Key Features

• Unifies ripgrep, BM25, and vector search in one interface
• Semantic search by meaning, with exact text and regex verification
• Multi-format search across source code, documents, and structured data
• Ranked, source-linked results with file and line locations
• Symbol-aware code indexing (symbols, signatures, breadcrumbs)
• Agent integrations: Codex, Claude Code, Qwen Code, Qoder, Cursor, GitHub Copilot, VS Code, OpenCode
• MCP tools for agents
• Local by default: files, indexes, and local models stay on your machine
• Remote embeddings only with your permission
• Choice of local or remote embedding models, with GPU/CPU and concurrency controls
• Direct mode or long-running local server mode
• Index stored in a .zvec-grep folder inside the project
• Runs on macOS, Linux, and Windows (Node.js 22+)

Useful Links