MassGen

Why we included this project

MassGen is for people who want higher-quality answers from frontier models without writing the orchestration layer themselves. You launch it in your terminal and it starts several agents on the same question in parallel, each one critiquing and building on the others' work across repeated rounds until they vote on the strongest answer. Since agents see anonymized outputs rather than who produced them, the final pick tends not to be swayed by the most confident or loudest voice in the room. The setup lives in a YAML file, so you can define your agent roster once, keep it in Git, and run the whole thing with a single command even if you don't write Python. That makes it a solid fit for teams that want reproducible multi-agent runs for research, analysis, or drafting, and a useful reference for how consensus-based coordination is actually built.