Claude-Mem (cmem.ai)

Claude-Mem (cmem.ai) is an open-source persistent memory system and context stream designed specifically for terminal coding agents and AI workflows, such as Claude Code, OpenClaw, and cursor-based environments. It captures tool executions, user decisions, file edits, and architectural constraints across development sessions, compressing and injecting relevant memories back into future agent turns.

graph LR
    subgraph "Coding Session"
        Terminal[CLI / OpenClaw / Claude Code] --> ToolExec[Tool Calls & Code Edits]
    end

    ToolExec --> Extraction[CMEM Observation Extractor]
    Extraction --> PrivacyFilter["Privacy Redactor (<private> Tags)"]
    PrivacyFilter --> LocalStore[(Local SQLite / Markdown Memory)]
    LocalStore <--> Sync[CMEM Cloud Sync]
    LocalStore --> PromptInjection[Context Injection into Next Agent Session]

Key Characteristics

• Agent Workflow Continuity: Eliminates the "cold start" penalty across development sessions by recording project conventions, debugging steps, and dependency decisions.
• Privacy Controls: Supports <private> markup tags in prompts and files to explicitly omit credentials, tokens, or sensitive context from persistent memory storage.
• Two-Tier Architecture (Local + Cloud):
  • Local Mode (OSS): Operates completely offline with local SQLite storage, zero external data leaks, and local embedding support.
  • CMEM Cloud (cmem.ai): Multi-device real-time sync, web viewer dashboard for memory stream inspection, and private Model Context Protocol (MCP) server endpoints.
• Native OpenClaw & Claude Code Support: Provides ready-to-run installation scripts (openclaw.sh, npx claude-mem install) and hook integrations for terminal agent frameworks.

Installation & Setup

Install locally via npx:

npx claude-mem install

Or configure via OpenClaw plugin:

curl -fsSL https://cmem.ai/openclaw.sh | bash

CLI Usage & Commands

Claude-Mem exposes simple commands to view, search, and manage stored context:

# Check memory health and stored record count
cmem status

# Search memory for past architectural decisions
cmem search "authentication architecture"

# Manually record a project rule or preference
cmem remember "Always run nix-shell before executing python tests"

Related Concepts

• LLM and Agent Memory - Overview of AI agent memory architectures.
• Memory Store (Julep) - Multi-tool cognitive memory platform with MCP support.
• Mem0 - Vector-first memory layer for assistants.
• Hermes Agent Memory Providers - Pluggable memory backends in Hermes Agent.