What problem does it solve? Understanding a large or polyglot codebase with grep-based exploration is slow and misses cross-service relationships. This Skill turns any codebase into a persistent knowledge graph so you can ask architecture, call-graph, and impact-analysis questions in plain English instead of manually tracing files. ## Core Features & Use Cases - Knowledge graph indexing: Parses 158 languages with tree-sitter plus semantic type resolution, exposing 15 MCP tools (index_repository, search_graph, trace_path, detect_changes, query_graph, get_architecture, and more) with CLI equivalents for scripting. - Cross-service route tracing: Matches HTTP routes to call sites with confidence scoring, plus gRPC/GraphQL/tRPC/Socket.IO detection, so a frontend fetch() call can link to the backend controller that serves it. - Staleness and verification guidance: Detects when uncommitted edits leave the graph stale (auto-watch is git-commit based, not filesystem based) and verifies graph answers before trusting them. - Use Case: Before modifying a shared API endpoint, ask the agent to trace every caller across your microservices and run detect_changes on your diff to see which symbols are affected and their risk classification. ## Quick Start Ask the agent to install codebase-memory-mcp with its one-line install script, restart the session, then say "index this project" and start asking architecture questions in plain English.
