osgrep

Semantic search lets you find code using natural language descriptions of functionality rather than exact text matches. osgrep performs semantic code search on local files, letting you query by behavior—such as 'where is authentication validated'—and returns matching code snippets with paths, line numbers, and relevance scores.

Yes. Semantic search is designed for architecture discovery, helping you locate distributed patterns across multiple files and directories. You can scope searches to specific directories for targeted investigations, making it effective for understanding how patterns like plugin loading or authentication span your codebase.

osgrep relies on a background live-indexing server to perform semantic search. You invoke queries with the --json flag to return structured results. The Skill works on local files without external dependencies, returning path, line, score, and content for each match.

Traditional grep matches exact text; semantic search understands conceptual meaning and behavior. osgrep finds code by what it does, not just the words in it, so 'how are plugins loaded' returns relevant code even if exact function names vary across your codebase.

Exactly. Semantic search is built for discovery when you understand only the concept or functionality. Describe the behavior you're investigating—validation logic, error handling, configuration patterns—and osgrep returns matching code across your entire codebase without requiring specific identifiers.