kitty

Codebase structure analysis is performed by parsing source files with tree-sitter to build a relationship graph, enabling dependency mapping and blast radius analysis across Python, TypeScript, and JavaScript projects.

Graph-based code exploration works by using tree-sitter to parse syntax trees and storing relationships in a SQLite graph, allowing you to query definitions, imports, and connections to navigate large codebases.

Blast radius analysis is executed by traversing the relationship graph to trace dependency impacts, allowing you to understand the downstream effects of modifying imports and definitions before making changes.

MCP-based workflows are supported by exposing an MCP server for agent-driven analysis, orchestrating multi-agent pipelines to brainstorm, plan, work, and review features directly against the codegraph.

Semantic search across a codegraph improves discovery by using annotated summaries and tags applied to graph nodes, providing contextual structural matches that go beyond plain text search results.

Tree-sitter parsing and SQLite graph storage are needed when you require structural understanding of definitions and imports, enabling AST-powered queries and dependency mapping rather than simple text matching.