mixedbread-ai-mgrep

Overview

mgrep is a CLI-native semantic search tool that lets you search code, text, PDFs, and images using natural language queries. Unlike traditional grep which matches literal strings, mgrep understands meaning, supports multiple languages, and automatically indexes your repositories in the background so searches feel as fast and immediate as regular grep.

Key Features

• Natural language search — Ask questions like "where do we set up auth?" instead of crafting regex patterns, and get semantically relevant results across your codebase.
• Multimodal search — Searches across code, text, PDFs, and images in a unified interface, with audio and video support coming soon.
• Background indexing — The mgrep watch command performs an initial sync respecting .gitignore, then continuously keeps the index updated as files change.
• Built-in web search — Query the web alongside your local files using the --web flag, combining local and internet results in one tool.
• Agent-friendly design — Reduces token usage by 2x compared to raw file content while maintaining search quality, with quiet output and thoughtful defaults designed for both humans and AI agents.

Who is this for?

mgrep is for developers and AI coding agents who want fast, meaning-based search across their projects without writing complex regex. It is particularly valuable for teams working with large codebases where traditional text search falls short, and for AI agent workflows that need efficient, low-token-cost file discovery.