zvec-grep
A local workspace search tool combines semantic and exact-text search for terminal users and AI agents on macOS, Linux, and Windows.
zvec-grep searches code, documents, and structured data for developers and AI agents that need to find relevant passages without knowing the exact words or file location. You can search from the terminal or give a coding agent access to the same indexed workspace.
- Semantic and exact search: Combine meaning-based retrieval with BM25 relevance ranking, then check specific words, identifiers, or regular expressions through ripgrep.
- Source-linked results: Retrieve ranked passages with source locations, preserving code symbols and signatures or document sections so you can inspect the evidence in context.
- Agent integrations: Connect workspace search to Codex, Claude Code, Cursor, GitHub Copilot, or OpenCode through agent integrations and MCP.
It's open source under Apache 2.0 and runs on macOS, Linux, and Windows, either through individual commands or a persistent local server. Files, indexes, and local embedding models stay on your machine. Optional remote embedding services receive data only with your permission.
Embedding backends include llama.cpp, Transformers.js, and Potion/model2vec, with CPU execution available. Hardware needs depend on the model and runtime; GPU execution can run into memory limits. The TypeScript/Node.js implementation requires Node.js 22 or newer, and the repository also contains a separately developed Rust implementation.