Overview

agentmemory is a cutting-edge persistent memory solution for AI coding agents, designed to eliminate repetitive explanations and context loss. Built upon the robust iii engine, it ensures your AI assistant remembers everything from previous sessions, significantly boosting productivity and streamlining your development workflow. agentmemory integrates seamlessly with a wide array of popular coding agents, including Claude Code, Cursor, Gemini CLI, and more, providing a unified memory server that shares context across all your tools.

Key Features

• Unprecedented Recall: Achieve an impressive 95.2% retrieval accuracy (R@5) on challenging benchmarks like LongMemEval-S, ensuring your agent always accesses the most relevant information.
• Vastly Reduced Token Usage: Save up to 92% on token consumption compared to traditional methods, translating to significant cost savings and faster processing times.
• Universal Agent Compatibility: Works with any agent supporting MCP, hooks, or REST APIs. Integrates with over 50 MCP tools and 12 auto-hooks for effortless setup.
• Advanced Memory Architecture: Leverages a hybrid search approach combining BM25, vector embeddings, and knowledge graphs for superior retrieval, extended by confidence scoring, lifecycle management, and auto-forget mechanisms.
• Zero External Dependencies: Operates with just SQLite and the iii-engine, eliminating the need for complex external databases.
• Real-time Memory Visualization: Monitor and replay your agent's memory build-up in real-time via an intuitive web viewer at http://localhost:3113.
• Effortless Installation & Management: Get started in 30 seconds with a single npx command. Includes built-in tools for demoing, importing sessions, and upgrading.

Typical Use Cases

• Eliminate Context Loss: Never re-explain your project setup, preferences, or common coding patterns. agentmemory ensures your agent retains this knowledge session after session.
• Accelerate Debugging: Quickly recall past bug fixes, error messages, and their resolutions, speeding up the debugging process.
• Streamline Feature Development: Maintain context on ongoing feature implementation, including design decisions, dependencies, and requirements, without constant re-briefing.
• Enhance Code Review: Agents with persistent memory can recall code quality standards and previous feedback, leading to more consistent and effective reviews.
• Centralized Knowledge Base: Utilize a single memory server across multiple AI coding tools, creating a cohesive and ever-growing knowledge base for your development environment.
• Cost-Effective AI Usage: Significantly reduce LLM API costs by intelligently injecting only the most relevant context, rather than sending large chunks of historical data.