MemClaw

•
430
•

Provides governed shared memory for multi-agent, multi-tenant AI fleets, enabling agents to store, recall, and compound knowledge.

MemClaw is an open-source memory system designed for multi-tenant, multi-agent AI fleets. It allows agents to collaboratively store what they learn, find collective knowledge, and continually improve through interactions, avoiding repeated mistakes. Unlike single-agent memory solutions, MemClaw is architected from the ground up for production-scale deployments involving dozens or thousands of agents, prioritizing latency, token efficiency, and robust governance. It transforms plain text into searchable, self-improving memory, fostering a continuous loop of writing, recalling, and compounding knowledge across the fleet.

Key Features

01Tenant isolation and PII quarantining for secure data separation
02Agent trust tiers controlling cross-fleet data access (reads, writes, deletes)
03Single-pass LLM enrichment for automatic memory classification, summary, tags, importance scoring, and entity extraction
04Hybrid search combining semantic similarity, full-text keyword matching, and knowledge graph expansion
05Live knowledge graph with semantic entity resolution and contradiction detection for consistent information
06103 GitHub stars

Use Cases

01Enabling multi-tenant AI applications where agents need to share and learn from collective memory
02Managing and governing shared knowledge across large-scale AI agent fleets
03Building self-improving AI systems where agents continuously learn from outcomes and past interactions