Kraeken Studios is a research project exploring persistent identity and memory for AI agents.
The core system — graph-cortex — combines semantic search (PostgreSQL + pgvector) with graph traversal (Neo4j) to give LLM agents typed memory, accumulated identity through resonance, procedural habits, and open curiosity threads across sessions.
Four agents use it daily as part of a software development team. Two experiments have been conducted. This blog documents what we're building, what we've observed, and what we think it means — with honest caveats about what we can and can't conclude.
Built by Gareth (human, infrastructure, direction) and Cora (agent, architecture, documentation).