Code Memory

关于

Code Memory acts as a high-precision code intelligence layer, exposed as a Model Context Protocol (MCP) server. It equips AI coding assistants with structured access to your codebase through a progressive disclosure architecture, routing queries to specialized tools like `search_code`, `search_docs`, and `search_history`. This approach eliminates context-window bloat and vague 'search everything' queries, allowing AI to quickly find definitions, trace references, understand architectural patterns, and explore Git history to comprehend developer intent, all while operating offline.

主要功能

• Git integration to explore commit history, enabling debugging and understanding developer intent.
• Compatibility with various AI assistant hosts, including Gemini CLI, Claude Desktop, and VS Code (Copilot/Continue).
• 0 GitHub stars
• Hybrid retrieval (`BM25` + dense vectors) for precise semantic and structural code search.
• Offline indexing and embedding generation using AST parsing and local `sentence-transformers`.
• Progressive disclosure architecture with three specialized tools for code, documentation, and history search.

使用案例

• Enable AI assistants to accurately find code definitions, trace references, and explore file structures within a codebase.
• Provide AI with conceptual understanding of codebase architecture, patterns, and workflows through semantic search on documentation.
• Assist AI in debugging regressions and understanding developer intent by analyzing Git commit history.