RBKunnela/ALMA-memory
Persistent memory for AI agents - Learn, remember, improve. Alternative to Mem0 with scoped learning, anti-patterns, multi-agent sharing, and MCP integration.
AI agents often forget previous interactions and learnings, leading to repetitive instructions and wasted effort. This project provides a persistent memory layer for AI agents, allowing them to remember past successes, failures, and preferences. It takes conversational history and task outcomes as input, and outputs refined strategies and knowledge, making future AI interactions smarter and more efficient for anyone developing or managing AI agents.
Used by 1 other package. Available on PyPI.
Use this if you need your AI agents to consistently learn, adapt, and improve over time, remembering specific strategies, outcomes, and anti-patterns across different sessions and even different AI platforms.
Not ideal if you only need short-term context within a single conversational session or if you prefer to rely solely on the built-in memory features of individual large language models.
Stars
20
Forks
4
Language
Python
License
—
Category
Last pushed
Mar 03, 2026
Commits (30d)
0
Dependencies
2
Reverse dependents
1
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/vector-db/RBKunnela/ALMA-memory"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
topoteretes/cognee
Knowledge Engine for AI Agent Memory in 6 lines of code
CaviraOSS/OpenMemory
Local persistent memory store for LLM applications including claude desktop, github copilot,...
verygoodplugins/automem
AutoMem is a graph-vector memory service that gives AI assistants durable, relational memory:
CortexReach/memory-lancedb-pro
Enhanced LanceDB memory plugin for OpenClaw — Hybrid Retrieval (Vector + BM25), Cross-Encoder...
divagr18/memlayer
Plug-and-play memory for LLMs in 3 lines of code. Add persistent, intelligent, human-like memory...