Definition
Plain language
A framework for building systems of multiple AI agents wired together as a graph.
As stated in the literature
A multi-agent orchestration framework representing agent collaboration as an optimizable graph; cited alongside Mixture-of-Agents as prior work whose fixed aggregator role contrasts with adaptive orchestration.
Why it matters: It treats teams of agents as a structure you can tune and optimize, making it easier to build multi-agent systems that work together effectively.
For example, you could use it to connect a research agent, a writing agent, and a fact-checking agent into one pipeline that passes work between them.
Heard on the show
“The prior work in this space — Mixture-of-Agents, GPTSwarm — uses one fixed model as the aggregator, which bottlenecks the whole system on tasks outside that model's strength.”Episode 166 — A Router That Beats the Frontier Models It Calls