Graph-based framework for LLM-based agents providing two high-level features: building agents from graphs and enabling customized automatic self-organization of agent swarms with self-improvement capabilities. Unifies diverse prompt engineering techniques by describing LLM agents as computational graphs. Each node implements a function to process multimodal data or query LLMs. Each edge describes information flow. Features automatic graph optimizers for node-level prompt refinement and agent orchestration improvement through graph connectivity changes.