arXiv cs.AISaturday · May 23, 2026FREE

AgentCo-op: Retrieval-Based Synthesis of Interoperable Multi-Agent Workflows

agentsworkflowsgenomicsretrieval

AgentCo-op, introduced in a preprint on arXiv (2605.20425v1), addresses the challenge of designing multi-agent workflows in open-ended scientific settings where tasks lack curated training sets, reliable metrics, and standardized interfaces. The framework retrieves and composes reusable skills, tools, and external agents into executable workflows through typed artifact handoffs. When execution evidence indicates failure, it applies bounded self-guided local repair to implicated components. In two open-world genomics case studies, AgentCo-op composed independently developed scientific agents and external tool repositories into auditable workflows without redesigning them or running global topology search. It coordinated specialized agents for spatial transcriptomics and gene-set interpretation to enable collaborative discovery from spatial transcriptomics data, and built a parallel workflow for cross-modality marker analysis on single-cell multiome data. AgentCo-op can also import a searched workflow as a structural prior and improve it by grounding nodes with retrieved components and applying local repair. The framework is designed for open-ended scientific settings, making it particularly relevant for researchers building complex analysis pipelines.

// why it matters

Enables developers to compose multi-agent workflows without redesign, using retrieval and local repair.

Sources

Primary · arXiv cs.AI
▸ Read original at arxiv.org

Like this? Get the next digest.