Claude-Mem: Persistent Memory System for AI Agents

2026-05-14 · source · source ↗ · raw capture · wiki page

Claude Code plugin that automatically captures everything Claude does during coding sessions, compresses it with AI (using Claude's agent-sdk), and injects relevant context back into future sessions.

Key Features

Persistent Context: Memory survives across sessions
Progressive Disclosure: Layered retrieval with visible token cost
Skill-Based Search: Natural language queries via mem-search skill
Web Viewer UI: Real-time memory stream at localhost:37777
Privacy Controls: <private> tags exclude sensitive content
Citation System: Reference past observations with IDs

Architecture

5 Lifecycle Hooks: SessionStart, UserPromptSubmit, PostToolUse, Stop, SessionEnd
Worker Service on port 37777 (Bun runtime)
SQLite for sessions/observations
Chroma Vector Database for hybrid semantic + keyword search
MCP Search Tools: 3-layer workflow (search -> timeline -> get_observations)

3-Layer Search (10x token reduction)

Layer 1: Compact index (~50-100 tokens)
Layer 2: Chronological context
Layer 3: Full observation details (500-1,000 tokens) only for filtered results

Requirements: Node.js 18+, Claude Code with plugin support, Bun runtime, SQLite 3. Apache 2.0 license.