agentUniverse is a Python-based multi-agent framework built to enable development of multi-agent applications using LLMs. It provides PEER/DOE collaboration models, support for various embeddings and LLM integrations, and observability; latest release adds AWS Bedrock, Ollama embedding, Jina rerank, and vector DB components.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
agentUniverse is a multi-agent framework based on large language models. It provides flexible and easily extensible capabilities for building individual agents, and a set of multi-agent collaborative pattern components that allow agents to perform duties and cooperate to solve problems. It includes integration capabilities for domain experience to inject expert knowledge.
How it works
The framework exposes multi-agent collaboration models such as PEER (Plan, Execute, Express, Review) and DOE (Data-finding, Opinion-inject, Express). It integrates with various LLM providers and embedding components, and supports adding domain knowledge through prompts, knowledge construction, and management. The repository describes components and patterns used to orchestrate multiple agents and their tools.
Getting started
Installation
pip install agentUniverse
Quick start notes
- Run the first example via the provided documentation. The README points to a Run Your First Tutorial Example for detailed steps.
- A standard project scaffold and quick guides exist for building single and multi-agent applications, including templates and MCP server usage. See the Quick Start and Get Started sections in the repository.
Visual workflow platform
pip install magent-ui ruamel.yaml
One-click Run is available by executing the sample application bootstrap file in the workflow platform example. More details are in the Product Platform Quick Start and Advanced Guide in the docs.
Recent releases
- v0.0.19 (2025-11-17): Added AWS Bedrock model support; added ollama embedding component support; added jina rerank component support; added vector database component support (Fais).
- v0.0.18 (2025-07-10): Added observability capabilities for agent applications based on the OTel protocol; established observation standards for aU agent applications based on OTel.
- v0.0.17 (2025-05-22): MCP Integration & Publishing Capabilities enabling rapid integration with MCPServer for agent, and allows publishing tools/toolkits from aU as MCPServer services.
- v0.0.16 (2025-04-17): New Tool Plugins including OpenCV OCR-based image text extraction, Shell command status query and execution tool, and universa.
- v0.0.15 (2025-03-03): Added new embedding components (Azure-OpenAI, Gemini, Doubao) and new knowledge loading reader component.
Traction
stars_7d: 0 stars_1d: 0
Behind the repo
This repository is associated with Ant Group for enterprise-scale intelligent agents, and describes a commercial product base connected to domain-expert workflows.
Caveats
License: Apache-2.0. Created: 2024-04-23. Last push: 2026-07-28. Language: Python. Primary language: Python. Open issues: 240. Stars: 2317. Forks: 416.