Ai Agent Orchestrator

This gives you the building blocks for coordinating multiple AI agents that can hand work between each other. You get a base agent class with memory, specialized agents for different roles (research, writing, review), and two orchestration patterns: a simple sequential flow and a supervisor mode where an LLM decides which agent runs next. The shared context system lets agents build on each other's outputs, and there's handoff logic so agents can pass tasks along. It's basically the plumbing you need when one LLM call isn't enough and you want to chain different specialized prompts together with some coordination logic. The supervisor pattern is the more interesting piece since it can adapt the workflow based on results.