Yuanzhu Technology

Yuan Agent-Turbo

From intent understanding and task planning to tool use and multi-agent collaboration, powering a more proactive way to work digitally.

Yuan Agent-Turbo is an agent model for AI agent development, task planning and multi-agent collaboration. It combines the understanding and generation abilities of large language models with task decomposition, tool calling, flow control and result consolidation, to build intelligent applications that carry out multi-step work toward a goal.

Compared with traditional interaction that answers one question at a time, AI agents focus on the whole task: understanding the goal, identifying the steps, choosing suitable tools, handling results and moving the task forward according to predefined logic.

Yuan Agent-Turbo extends this approach to multi-agent collaboration, letting agents with different responsibilities divide the work around a shared goal and offering a more flexible way to handle research and analysis, business processing, knowledge work and automated office tasks.

Every capability serves a real task.

  • Powerful agent buildingBuilt on large language models and task planning, it supports designing AI agents around different task goals. Each agent can be given responsibilities, inputs, objectives and execution rules according to real needs, so it understands the task and chooses a suitable approach within a defined scope. With clear task design, developers can build agent applications for research, analysis, content processing or business assistance.
  • Autonomous task planning and decompositionFaced with a complex goal, it breaks the overall task into sub-tasks that are easier to execute and manage. By defining steps, dependencies and intermediate results, it helps agents complete the work step by step according to plan, rather than giving a single answer. For tasks that span many stages, this creates a clearer execution path and more flexible task organization.
  • Tool calling and external capability extensionIt plans calls to external tools, knowledge bases and other execution capabilities according to task needs. By connecting suitable external resources, agents can bring information retrieval, content processing and other supported operations into the task flow, extending what a single model can do.
  • Flexible multi-agent collaborationMultiple agents with different responsibilities can be configured to complete more complex work through task assignment, information exchange and result consolidation. For example, in a research task, separate agents can collect materials, analyze content, organize viewpoints and write the report, while a coordination module brings the results together. This organizes complex tasks into units with clear responsibilities, so roles and execution flows can be adjusted to real needs.
  • Task execution and process coordinationWork is organized around task status, execution steps and intermediate results, and moves forward according to preset rules. For work that takes several steps, tool calls and flow control arrange the execution order and lead into the right follow-up step based on each result. For important business operations, appropriate permission controls and human confirmation keep execution in line with the expected rules.
  • An extensible agent application systemAgents with different responsibilities can be designed around business needs, with new tools, task types and collaboration patterns added over time. A modular approach to organizing agents helps developers build applications at every level, from single-task assistants to multi-role collaborative systems.

Yuan Agent-Turbo extends AI from single interactions to task planning, tool use and multi-agent collaboration, helping users build more flexible, goal-driven intelligent applications.