Memory Masterclass: Make Your AI Agents Remember What They Do! — Mark Bain, AIUS
Mark Bain · Vasilije Markovic · Daniel Chalef · Alex Gilmore
AI Engineer World's Fair 2025 · 51:25
AI agent memory infrastructure
Cognee builds open-source memory infrastructure that lets AI agents retain knowledge across sessions and share it across workflows. Developers use its Python and TypeScript SDKs to remember information, recall relevant context, improve memory through feedback and forget unwanted data. Cognee Cloud provides managed infrastructure; teams can also self-host or use its Rust core for edge environments. Applications include coding assistants, sales deal memory and scientific research: Bayer uses Cognee to connect research knowledge and generate hypotheses.
CEO and founder Vasilije Markovic started Cognee with his team in Berlin in 2024. Its engine combines knowledge-graph relationships, vector similarity and relational storage for documents and provenance. Feedback can refine graph connections and remove stale information. A 2025 research paper studied how tuning chunking, graph construction, retrieval and prompting affects multi-hop question answering, finding gains that varied across datasets and metrics. This work treats memory as an evolving structure, rather than simply an accumulation of retrieved text.
In June 2026, the company reported approximately 6 million memories created monthly across more than 100 companies. It had raised a $7.5 million seed round led by Pebblebed earlier that year, with participation from 42CAP and Vermilion Ventures. By August 2026, Cognee reported crossing 30,000 GitHub stars.
Mark Bain · Vasilije Markovic · Daniel Chalef · Alex Gilmore
AI Engineer World's Fair 2025 · 51:25
Affiliations reflect their AIE appearances, not necessarily current employment.
The workshop moves into practical integration through memory-server demonstrations with Claude Desktop. Bain compares MCP integrations across Neo4j, Graphiti, Cognee, and Mem0.
Affiliations reflect each recorded session, not necessarily current employment. This is a shared workshop led by Mark Bain, with guest presenters Vasilije Markovic, Alex Gilmore, and Daniel Chalef.