OPEN LINK ↗
// 76d agoOPENSOURCE RELEASE
Memorine brings local memory to agents
Memorine is a new open-source Python memory system for AI agents that stores facts, events, procedures, and shared knowledge in SQLite instead of relying on cloud services or a vector database by default. It adds decay-based recall, contradiction detection, causal event chains, optional embeddings, and an MCP server in a lightweight package that feels aimed squarely at local-agent builders.
// ANALYSIS
This is a smart swing at one of the weakest parts of agent stacks: most “memory” products are retrieval layers, not actual long-term memory models.
- –The core pitch is strong because it stays local and dependency-light: pure Python, SQLite, zero required external services
- –Decay curves, contradiction checks, and procedure tracking make it more interesting than a basic notes store or RAG cache
- –Optional embeddings keep the base install simple while still giving power users semantic recall when they need it
- –MCP support makes it practical for the current agent tooling ecosystem, especially for local and self-hosted workflows
// TAGS
memorineagentopen-sourceself-hostedmcpdevtool
DISCOVERED
76d ago
2026-03-11
PUBLISHED
78d ago
2026-03-10
RELEVANCE
8/ 10
AUTHOR
coloresmusic