claudemem-search

Semantic code search uses vector embeddings and BM25 ranking to find relevant code by meaning, not just keyword matches. claudemem-search analyzes your codebase with tree-sitter AST parsing and symbol graphs to return contextually ranked results across multi-language repositories, enabling faster debugging and comprehension.

AST-based structural analysis parses code into abstract syntax trees, mapping symbols, function calls, and dependencies automatically. claudemem-search constructs symbol graphs and applies PageRank ranking to show you which functions are most critical and how they connect, eliminating manual navigation.

The callers and callees commands trace function relationships through AST parsing and symbol graph construction. claudemem-search identifies every place a function is invoked and everything it calls, supporting refactoring, impact analysis, and dependency debugging across your codebase.

Start with the map command to generate structural overviews using AST graphs and PageRank-ranked symbols. claudemem-search reveals architecture, key entry points, and symbol importance before you dive into reading, accelerating onboarding and comprehension of new codebases.

AST analysis with tree-sitter supports multiple languages in a single repository. claudemem-search handles language-agnostic structural parsing and symbol graphs, letting you search and navigate across polyglot codebases without switching tools.

Context-aware symbol queries combine vector embeddings with BM25 ranking and AST-based symbol graphs. claudemem-search returns ranked results filtered by symbol type and relationship, delivering precise hits instead of noise, with version-aware fallbacks for reliability.