Mistral AI has released version 0.9 of its open-source Multi-Agent Framework, introducing an autonomous negotiation mechanism that allows agents to re-decompose and reassign tasks when a teammate hits a capability gap, with zero human intervention required. The update significantly improves success rates on complex, long-horizon tasks and signals a major shift in multi-agent AI collaboration—from manual orchestration to genuine team autonomy.

Mistral AI recently unveiled version 0.9 of its open-source Multi-Agent Framework. The centerpiece of this release is the autonomous negotiation mechanism: when any agent in the team is unable to complete its assigned task, the system automatically re-decomposes the task and redistributes it to other agents whose capabilities match the requirements. The entire process runs without human oversight, and the company says this capability can markedly improve completion rates for complex, long-running tasks.

he rapid iteration of multi-agent frameworks—built on increasingly capable large language models