TL;DR
agentmemory is a persistent memory framework that helps AI coding agents maintain context across sessions, supporting 53 MCP tools and hybrid search.
Key features
Persistent memory: Agents remember previous work, eliminating the need for re-explanation
Hybrid search: Combines semantic and keyword search for high accuracy
Confidence scores: Assigns confidence to memory items for importance-based management
Lifecycle management: Automates memory creation, update, expiration, and deletion
Knowledge graph: Represents relationships between memories as a graph to support reasoning
53 MCP tools: Rich toolset for integration with various agents
Benchmark-based: Achieves 95.2% retrieval recall and 92% token savings based on real-world benchmarks
When to use it
When coding agents like Claude Code, Copilot CLI, or Cursor need long-term memory
When agents must retain and utilize project context, user preferences, and past decisions
To maintain consistency across complex coding tasks spanning multiple sessions
To add memory capabilities to any agent supporting the MCP protocol