Show HN: A local-first memory store for LLM agents (SQLite)
6 days ago
- #AI Memory
- #Self-Hosted
- #OpenSource
- OpenMemory is a long-term memory solution for AI systems, offering self-hosted, local-first, and scalable features.
- It simplifies AI memory integration with just 3 lines of code, contrasting traditional vector databases that require extensive setup.
- OpenMemory provides features like persistent memory, multi-sector cognitive structure, natural decay, and graph-based recall.
- It supports various deployment modes including Node.js SDK, Python SDK, and Backend Server.
- OpenMemory includes a VS Code extension for tracking coding history and integrates with MCP (Model Context Protocol) for AI tools.
- The system offers temporal memory capabilities, allowing agents to reason about changing facts over time.
- OpenMemory uses Hierarchical Memory Decomposition for input processing and memory recall.
- Performance metrics include 115ms avg recall @100k, 338 QPS throughput, and 95% accuracy@5.
- Security features include AES-GCM encryption, API keys, user isolation, and no telemetry unless allowed.
- Future roadmap includes learned sector classifier, federated memory clusters, and memory-visualizer 2.0.