MassGen

MassGen is a multi-agent system in which one task goes to several AI agents that work in parallel, observe each other's progress and refine their answers until they converge. The README says it started from the "threads of thought" and iterative refinement ideas in an AG2 article on reasoning, and extends classic multi-agent conversation. Its feature table lists cross-model synergy, parallel processing, intelligence sharing, consensus building and live visualisation of what each agent is doing.

Agents can be backed by many providers. The README names OpenAI (GPT-5.x series, GPT-4.1, o4-mini), Azure OpenAI, Claude and Claude Code, Gemini, and fast or open-model hosts such as Cerebras, Together, Fireworks, Groq, OpenRouter, Qwen, Kimi and Zhipu, plus tools including web search, code execution and MCP. You can start with a single agent, then move to a multi-agent configuration, and there are modes for MCP and file-system operations in a managed workspace. It also provides an `AI_USAGE.md` guide for running MassGen from inside an LLM agent.