Store Your AI Agent's Memory and Context in a Knowledge Graph
summary
Runs OpenKlo with a memory adapter called Cogny adapter and uses Falco DB (RedisGraph descendant) to store agent memory as nodes and edges, not flat files. Demonstrates two agents, Liz and Will, adding company and person nodes in Tokyo and querying one‑ and two‑hop relationships in milliseconds. Claims hybrid vector + graph search, per‑agent 1,000 agents / 10,000 graphs scale with low token cost, live demo recorded and working end to end.
AI-written summary · about
chapters (12)
- 0:00Agent amnesia and context
- 2:14FalcoDB as graph database
- 5:02Short-term agent memory problems
- 6:48Graph databases for agents
- 9:16Demo setup with OpenKlo agents
- 12:26Adding Tokyo company data
- 16:58Q&A on nodes and embeddings
- 18:00Stack configuration and tokens
- 20:26Token management and cost
- 21:50FalcoDB open source and scalability
- 23:57Graph vs vector databases
- 25:00Q&A on multi-graph scalability
related talks
29mThe AI Agent Memory Landscape
38mSearch is Back: Solving the "Context Crisis" for AI Agents
51mMemory Masterclass: Make Your AI Agents Remember What They Do!
21mBuilding memory for AI Agents
01h 34mHands-on AI workshop: Graph RAG, Memory & Multimodal Agents
15mBuilding Evolving AI Agents Via Dynamic Memory Representations Using Temporal Knowledge Graphs
21mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
12mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval