project-memory

Store architectural decisions, patterns, and learnings using mcp__memory__store with tags like "architecture" or "decisions." This creates searchable project memory that survives session restarts, eliminating repeated explanations of the same choices to team members returning to the codebase.

Yes. Use mcp__memory__search to query stored memories by content, tags, and metadata. The Skill surfaces relevant architectural decisions, solved problems, and implementation patterns when you encounter similar technical challenges, reducing rework and decision fatigue.

Store project-specific memories with consistent tagging and metadata (project, type, topic). Team members searching the same tags retrieve shared context across sprints, ensuring everyone recalls the same rationale for auth systems, API patterns, database choices, and other critical decisions.

Include content describing the decision or pattern, tags categorizing the memory (e.g., "auth", "performance", "database"), and metadata fields like project name and type. This structure enables precise search retrieval when you need context during code reviews, refactoring, or onboarding.

Use project memory for cross-cutting decisions, architectural rationale, and learnings that apply across multiple files or sprints—why JWT was chosen over sessions, for example. Reserve inline comments for local implementation details. Memory persists searchably; comments live in code and scatter across the repository.