Microsoft Research has open-sourced Agensh, a system for coordinating groups of AI agents without a central controller assigning each task. The agents share a Git repository, chat and a board for findings, and can organize their work among themselves.
Microsoft says Agensh scales from one to 1,024 agents on ProgramBench, with a six-hour time budget. The announcement does not give performance results or explain which models were used, so the scale figure alone does not show how well the system performs.
Why it matters
For developers exploring multi-agent software, Agensh offers an open-source coordination setup to examine, but its practical performance is unclear without reported results or model details.
