Codebase Memory transforms your project's source code into a detailed, persistent knowledge graph, offering a highly efficient alternative to traditional file-by-file exploration. It meticulously parses 35 programming languages to extract functions, classes, modules, call relationships, and cross-service HTTP links using tree-sitter. This structural analysis backend drastically reduces token consumption for AI assistants by providing precise, context-rich results through Cypher-like queries, seamlessly integrating with MCP-compatible clients like Claude Code, Cursor, and Codex CLI, or via its own powerful CLI mode.
Key Features
01Supports 35 programming languages with tree-sitter parsing.
02Generates comprehensive architecture overviews and call graphs.
03Maps Git diffs to affected symbols with risk classification.
04Achieves 99.2% token reduction compared to file-by-file exploration.
05Offers sub-millisecond graph queries and incremental re-indexing.
0683 GitHub stars
