From Target Molecule to Simulation-Ready Digital Microbial Life

Engineering microorganisms to produce target molecules is a central goal of synthetic biology, yet designing a simulation-ready microbial system remains labor-intensive and highly dependent on expert knowledge. Although recent large language model (LLM)-based scientific agents can automate parts of this workflow, they rarely produce executable designs that can be directly used for downstream computational analyzes. To address this challenge, We introduce Origo, a multi-agent system that autonomously constructs simulation-ready digital microbial life from a target molecule. Origo combines an expert-designed primitive task pool, task-scoped tool binding, and over 50 specialized MCP tools to provide structured task decomposition and stage-specific biochemical capabilities throughout the design process. Starting from a target molecule, Origo can autonomously plan biosynthetic pathways, builds genome-scale metabolic models (GEMs), identifies candidate enzymes for individual reaction steps, and compiles the resulting designs into models ready for flux balance analysis (FBA). Experimental results on end-to-end generation of GEMs and reaction-level enzyme mining show that Origo outperforms general-purpose LLMs and other agent-based baselines, producing models and enzyme candidates that are more executable and biochemically plausible. By delivering simulation-ready in silico biological designs rather than reaction lists alone, Origo provides a practical starting point for further scientific reasoning, metabolic engineering, and hypothesis generation for wet-lab validation.