- CollectionMemory & State
- Classificationagent-native
- LicenseOpen-source
- MCPUnavailable
- Agent SkillUnavailable
- InterfacesREST · SDK · API
What it does
memU is a persistent memory framework specifically built for 24/7 proactive AI agents — agents that run continuously, monitor data streams, and act without explicit user commands. The critical insight: most agent memory systems are optimized for interactive sessions, but proactive agents need memory that works at millisecond latency while the agent is idle, only invoking expensive LLM reasoning when something truly relevant is detected.
memU solves this with dual-mode retrieval:
- Fast Context (RAG mode) — embedding-based monitoring running continuously, sub-millisecond latency, near-zero cost
- Deep Reasoning (LLM mode) — full LLM capabilities activated only when Fast Context detects a relevant pattern
The agent saves ~90% of token costs because expensive LLM reasoning only fires when the lightweight monitor says “this matters.”
Agent-facing interfaces
- Interaction pattern: SDK / REST
- Catalog protocols: REST, SDK, API
- MCP: ⚠️ Not yet published as standalone MCP server.
- Agent Skill: ⚠️ No official skill published yet.
- Onboarding type:
sdk
How agents onboard
Interaction pattern: SDK / REST
pip install memu
# Initialize memory layer
from memu import MemU
memory = MemU(agent_id="my-agent")
# Continuous monitoring mode — for proactive 24/7 agents
memory.monitor(data_stream=my_stream, callback=on_relevant_event)
# Or: standard add/retrieve
memory.add(content="User has deadline on Friday")
context = memory.retrieve("user schedule and priorities")
See the full onboarding section in the dossier.
Protocol surface
| Interface | Detail |
|---|---|
| Python SDK | pip install memu |
| REST API | Memory CRUD, stream monitoring, context retrieval |
| OpenClaw Integration | Native plugin for OpenClaw memory sessions |