Agent Memory

•
3
•

Unifies long-term memory across diverse AI agents and development environments through a self-hosted service.

Agent Memory offers a crucial solution for developers and teams using multiple AI coding agents by consolidating their disparate memories into a single, continuously maintained service. Instead of agents like Claude Code, Codex, or Cursor starting from scratch with each interaction or tool switch, Agent Memory provides an authoritative, shared memory plane. This self-hosted layer automatically captures durable facts, manages context lifecycle with features like decay and reinforcement, and enables cross-tool recall, ensuring consistency and reducing repetitive context setting across your entire AI-assisted workflow.

Características Principales

01Self-hosted deployment options for privacy and control (local, homelab, LAN, cloud VM)
02Flexible integration via Streamable HTTP MCP for agents and REST API for broader integrations
03Automatic context management including smart extraction, deduplication, decay, and reinforcement
042 GitHub stars
05Shared long-term memory across multiple AI agents (e.g., Claude Code, Codex, Cursor, Gemini CLI, VS Code, OpenClaw)
06Durable memory persistence using LanceDB with hybrid semantic and keyword retrieval

Casos de Uso

01Developers and small teams who frequently switch between multiple AI coding agents and desire a unified memory
02Users who prefer self-hosting their AI infrastructure for enhanced privacy and control over their data
03Individuals seeking to establish a shared, automatic memory layer for AI agents without the complexity of managing a knowledge graph