EvoAgentX

Why we included this project

EvoAgentX starts from a single goal description, builds a structured multi-agent workflow around it, scores it with built-in evaluators, and iterates to improve the result. You describe the task, the framework drafts the pipeline, and you can keep the parts you want while it gets better with each pass. It also ships the practical pieces most projects need anyway: short- and long-term memory, a library of tools for real-world actions, and human-in-the-loop checkpoints that let a person review or correct a workflow mid-run. That makes it a solid fit for researchers studying self-optimizing agents and for small teams prototyping agentic systems without hand-writing the orchestration glue.