An AI that takes notes on other AI's work in real-time
Transform ephemeral AI conversations in REAL-TIME as a permanent, searchable archive. Visualize development timelines, track decisions across commits, and collaborate with your team.
Claude-Mem uses progressive disclosure, starting sessions with lightweight indexes (titles, types, timestamps) and fetching full observations only on demand. This keeps token efficiency high by default while preserving access to complete context when the AI needs depth.
Claude-Mem records before-and-after context for every observation, allowing users to see why decisions led to specific outcomes. Auto-categorization by decision, bugfix, feature, and discovery enables precise filtering and understanding of development causality.
Claude-Mem is developing RAD (Real-time Agent Data), an open standard for AI agent memory with hook-based architecture, intelligent compression, and temporal awareness. This positions users to adopt a future-proof memory standard as it becomes available.
Claude-Mem supports dual-mode querying: file-path scoping (e.g., 'decisions for src/auth/index.ts') and semantic concept scoping (e.g., 'what do we know about auth'). This enables surgical precision in retrieving relevant historical context.
Claude-Mem automatically captures and archives AI assistant work through live observation, creating searchable records of decisions, bug fixes, and architectural choices. Users can query previous sessions by time, file, or concept, eliminating the need to re-explain context to the AI in new sessions.