codebase-memory-reference

You query a codebase knowledge graph using Cypher-like patterns with search_graph and query_graph to traverse edge types like CALLS and IMPORTS. This returns contextual code retrieval results by navigating defined node relationships across indexed repositories.

Codebase memory graph edge types include CALLS, IMPORTS, DEFINES, and HANDLES for mapping semantic relationships. These edges connect node labels to represent structural dependencies and execution paths within the indexed repository knowledge graph.

You index a repository by invoking the index_repository tool, then monitor progress with index_status. Once indexing completes, the codebase memory graph supports search_graph queries and code retrieval across the indexed project structure.

Yes, the codebase memory MCP tool suite provides list_projects to enumerate indexed repositories and delete_project to remove outdated ones. This manages the knowledge graph scope to ensure precise graph queries across active codebases.

The best way to retrieve code in a large codebase is using query_graph with documented Cypher-like patterns to target specific node labels and edge types. This narrows search_graph results to precise contextual code segments based on structural relationships.

Graph queries return empty results when search_graph patterns are not aligned with the documented Tools list, Edge Types, Node Labels, and Cypher-like syntax. Matching the exact API set and graph primitives ensures contextual code retrieval from the indexed repository.