yoanbernabeu/grepai

What it solves

It replaces traditional text-based search (like grep) with semantic search for codebases. This allows developers to find relevant code by meaning and intent rather than relying on exact keyword matches or regular expressions, which is especially useful when naming conventions are inconsistent.

How it works

grepai uses vector embeddings to index code and understand its conceptual meaning. It can be configured to use local embedding providers like Ollama or LM Studio, or OpenAI, ensuring that code can remain 100% local for privacy. It includes a file watcher to keep the index updated automatically and can be integrated as an MCP server for AI agents.

Who it’s for

Developers who want a more intuitive way to navigate large codebases and users of AI coding agents (such as Claude Code, Cursor, and Windsurf) who want to reduce the input token count by providing only the most relevant context.

Highlights

• Semantic Search: Find code by intent (e.g., searching for "authentication logic" to find a specific session handler).
• Privacy-First: Supports fully local indexing and search.
• AI Agent Integration: Works as an MCP server, allowing agents to call the tool directly.
• Call Graph Tracing: Ability to trace who calls a specific function.
• Automatic Indexing: A background daemon keeps the codebase index fresh.