Agent Memory
Persists and retrieves repository-specific knowledge to maintain context across different AI coding sessions.
Agent Memory provides a standardized way for Claude to retain crucial details about a codebase by utilizing AGENTS.md files at the repository root. It automatically captures and stores repository structures, build commands, code styles, and custom workflows, ensuring that every future interaction starts with a deep understanding of your project's unique requirements. By acting as a persistent long-term memory, it eliminates the need to re-explain project basics, setup procedures, and development standards every time a new session begins.
主な機能
01User-controlled knowledge updates with explicit confirmation prompts
02Persistent storage of build, lint, and test commands
03Automated repository structure mapping and documentation
04Intelligent filtering to exclude temporary issue-specific data
0540 GitHub stars
06Automatic context injection from existing AGENTS.md files
ユースケース
01Maintaining a persistent source of truth for repository commands to prevent AI hallucinations
02Onboarding an AI agent to a complex codebase with non-standard build scripts
03Standardizing code style and workflow preferences for AI-assisted development