zvec-grep is a local-first search tool that unifies keyword search, BM25 ranking, and vector search behind a single interface for both human users and AI agents.
The tool addresses the problem of finding relevant information across workspaces when exact keywords are unknown or when meaning-based discovery is needed. It combines ripgrep for exact matching, BM25 for relevance ranking, and vector search for semantic understanding. Results are ranked by relevance and linked to source locations, reducing the number of searches and tokens required to find what you need. The approach keeps files and indexes local by default, with remote embeddings only used when explicitly permitted.
The tool suits developers and AI agents working with mixed content types including source code, documents, and structured data. It works across macOS, Linux, and Windows. The same indexed workspace can be searched from the command line or programmatically by agents, making it useful for both interactive exploration and automated workflows. The project positions itself as an alternative to traditional grep-based searching when semantic understanding or relevance ranking would be more effective than keyword matching alone.
The project maintains active continuous integration and demonstrates ongoing development across its codebase. The tool is distributed as an npm package and requires Node.js 22 or newer, indicating a commitment to modern JavaScript tooling standards.
