March 16, 2026AgentsOpen SourceCoding

claude-mem: Persistent Memory Plugin for Claude Code via Agent SDK

claude-mem is a Claude Code plugin that captures everything Claude does during coding sessions, compresses context using Claude's Agent SDK, and injects relevant information into future sessions — giving Claude Code persistent memory across conversations.

The plugin works through 5 lifecycle hooks (SessionStart → UserPromptSubmit → PostToolUse → Summary → SessionEnd) that observe every tool invocation, capture inputs and outputs, process them through Claude's Agent SDK for semantic compression, and store the results in a local SQLite database with full-text search. Previous session context automatically appears in new sessions after installation.

claude-mem introduces RAD (Retrieval Augmented Development), an open standard for AI agent memory that mirrors how RAG standardizes external knowledge retrieval. The architecture includes a worker service with an Express API on port 37777 and a search skill accessible via HTTP API.

The project has gained massive traction, trending on GitHub with over 36,000 stars. It's open source under AGPL-3.0.

GitHub: https://github.com/thedotmack/claude-mem | Website: https://claude-mem.ai/
← Previous
LangChain Deep Agents: A Batteries-Included Agent Harness for Complex Tasks
Next →
NanoClaw Partners with Docker to Sandbox AI Agents in MicroVMs
← Back to all articles