Spec Memory for Claude
Manages and searches project specification files to provide Claude with a persistent, context-aware memory of design decisions and requirements.
Spec Memory is a specialized tool for developers practicing spec-driven development, allowing Claude to index, search, and retrieve project specifications from Markdown files. It uses a local SQLite database with Full-Text Search (FTS5) to provide rapid access to design docs, requirements, and task lists without requiring external embedding APIs. By injecting only relevant chunks or a high-level index into the LLM's context, it efficiently manages token budgets while ensuring Claude stays aligned with the project's technical history, architectural goals, and past implementation decisions.
주요 기능
01Automated indexing with content-hash deduplication for high performance
02PreToolUse hook for automatic index injection at the start of AI sessions
0317 GitHub stars
04Full-text search (BM25) of project specs using local SQLite integration
05Support for both directory-based and flat-file specification layouts
06Versioned archiving for tracking the evolution of design documents
사용 사례
01Searching for prior art and existing implementation patterns within large codebases
02Recalling previous architectural decisions when starting a new feature branch
03Injecting specific technical context when handing off tasks to sub-agents