Bioinfoysis is a multi-agent system for long-horizon bioinformatics tasks. Unlike typical agents that produce final answers and discard intermediate steps, Bioinfoysis represents each request as a persistent, artifact-grounded run. A global planner creates an executable checklist, and after each worker step, structured handoffs bind results to specific agents and plan versions, preventing reuse of stale evidence. A controlled runtime validates generated scripts, tables, and figures before they enter downstream analysis.
Evaluated on BixBench and two LAB-Bench 2 tracks, Bioinfoysis achieved state-of-the-art accuracy of 82.4% on BixBench. Across four language models, average accuracy improved from 27.81% to 64.13% on SeqQA2 and from 3.13% to 31.25% on DbQA2. These results show that governing planning, memory, and evidence flow is critical for reliable bioinformatics automation.