agentuniverse-ai/agentUniverse

agentUniverse is a LLM multi-agent framework that allows developers to easily build multi-agent applications.

What it solves

agentUniverse tackles the complexity of building intelligent agents that must handle domain-specific tasks with high expertise. Instead of designing a single monolithic agent, it enables multiple agents to collaborate by following proven patterns, and it allows injecting domain knowledge into the agent workflow. This makes it easier to create agents that perform at a professional level, especially in fields like finance, where precision and structured reasoning matter.

How it works

agentUniverse provides a core set of components to build individual agents, plus a library of multi-agent collaboration patterns (called 'pattern components') that have been tested in real business scenarios. Two patterns are currently open-sourced:

  • PEER (Plan, Execute, Express, Review) – breaks complex problems into steps, executes them sequentially, and iterates based on feedback. It is designed for reasoning and analytical tasks like event interpretation or industry analysis.
  • DOE (Data-fining, Opinion-inject, Express) – focuses on data-intensive tasks that require high accuracy and integration of expert opinions, such as financial report generation.

The framework supports easy integration with many LLM vendors—including Qwen, Deepseek, OpenAI, Claude, Gemini, Llama, KIMI, WenXin, chatglm, BaiChuan, and Doubao—via simple configuration. It also offers tools, knowledge bases, RAG, memory, prompt management, and a visual agentic workflow platform. Observability is built-in using the OpenTelemetry protocol to track agents, LLMs, and tools.

Who it’s for

Developers and enterprises who want to build professional-grade AI agents without starting from scratch. It is particularly suited for domains with strict requirements, like finance, where the framework originated at Ant Group. The commercial product 'Zhi Xiao Zhu' for financial professionals is built on agentUniverse, so it is battle-tested in large-scale production.

Highlights

  • Provides two validated multi-agent collaboration patterns (PEER and DOE) that improve reasoning and analytical performance
  • Supports a wide range of LLM providers, including Qwen, Deepseek, OpenAI, Claude, and more, with simple configuration
  • Offers flexible tools, knowledge bases, RAG, memory, and prompt management to enhance single-agent and multi-agent capabilities
  • Includes a visual agentic workflow platform for creating and orchestrating workflows via a canvas interface
  • Features an OpenTelemetry-based observability framework for monitoring the full lifecycle of agents, LLMs, and tools
  • Originates from real-world financial business practices at Ant Group and powers the commercial AI assistant 'Zhi Xiao Zhu'

Related

  • Project
    nullpointexception-i/agent-sphere

    AgentSphere is a Java‑Spring/React platform that runs LLM‑driven ReAct loops, supports multi‑agent delegation, real‑browser automation via a Chrome extension, and a public skill marketplace. It integrates many model providers, offers OIDC SSO, and provides an embeddable chat widget.

  • Project
    AgentEra/Agently

    Agently is a Python framework that turns LLM calls into structured, observable, and restart‑safe runtime units. It provides contract‑first output schemas, streaming of partial results, tool/action integration, versioned skills, managed execution resources, and an event‑driven workflow engine (TriggerFlow) for building reliable AI services.

  • Project
    mims-harvard/ToolUniverse

    ToolUniverse is an open‑source platform that lets any LLM act as an AI scientist. It provides a standardized protocol, a catalog of 1 000+ scientific tools, a local MCP server, a CLI (`tu`), and a Python SDK for discovering, composing, and running tools—enabling agents like TxAgent and Medea to perform drug‑discovery and omics analyses.

  • Project
    Agent-Field/agentfield

    AgentField is an open-source control plane that transforms AI agent code into production-ready microservices, handling orchestration, scaling, and governance automatically.

  • Project
    microsoft/agent-framework

    Microsoft Agent Framework is an open‑source SDK for building production‑grade AI agents and multi‑agent workflows in Python, .NET, and Go. It offers provider‑agnostic LLM access, graph‑based orchestration, middleware, OpenTelemetry observability, declarative YAML agents, skill integration, and easy hosting on Azure or Microsoft Foundry.