Anus - An open-source AI agent project generated by Manus, replicating some of Manus's features

What is Anus

Anus (Autonomous Networked Utility System) is an open-source autonomous agent project generated by Manus , replicating some of Manus's features. Anus supports natural language instruction execution, multi-agent collaboration, web interaction, document processing, code execution, and multimodal input processing. Anus uses a hybrid architecture that combines the simplicity of a single agent with the power of multiple agents, supports OpenAI, open-source models, and local deployment, and provides a rich tool ecosystem and flexible model integration options. Anus aims to provide developers, researchers, and AI enthusiasts with a powerful, flexible, and easy-to-use tool to promote the widespread application and innovation of AI technology.

Main Features of Anus

• Task Automation: Executes complex tasks based on natural language instructions, supporting single-agent or multi-agent modes.
• Multi-Agent Collaboration: Supports multi-agent systems with predefined roles (such as researcher, analyst, writer) collaborating to complete complex tasks.
• Multimodal Input Processing: Supports multiple input forms such as text, images, and audio, enabling image recognition, audio transcription, and video analysis.
• Rich Tool Ecosystem：
  • Web Interaction: Supports web automation, data scraping, form filling, and authentication handling.
  • Document Processing: Supports PDF and Office document analysis, as well as OCR recognition.
  • Code Execution: Supports code generation and secure execution in languages such as Python.
• Flexible Model Integration: Supports OpenAI models, open-source models (such asLlama, Mistral), and local deployment.

Technical Principles of Anus

• Based on Manus's Generation Capabilities: The entire project's design, coding, and documentation were completed autonomously by Manus, which referred to existing knowledge and open-source projects on the internet during the generation process.
• Hybrid Agent Architecture: Combines the efficiency of a single agent with the collaborative capabilities of multiple agents, dynamically switching modes based on task complexity. In a multi-agent system, agents divide labor and collaborate based on predefined or custom roles, communicating and resolving conflicts through structured protocols.
• Dynamic Task Planning: Decomposes complex tasks into multiple subtasks and executes them step by step based on an intelligent planning system. Dynamically allocates computing resources based on task requirements to optimize performance.
• Tool Ecosystem: Integrates multiple tools (such as the web automation tool Playwright, document processing tools, code execution sandboxes, etc.) and extends functionality through a plugin system.

How to Install Anus

• Install via pip (recommended for users)。

# 安装Anus
pip install anus-ai

# 验证安装
anus --version

# 安装所有可选功能
pip install anus-ai[all]

git clone https://github.com/nikmcfly/ANUS.git
cd ANUS

python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

pip install -e .

• • Verify installation：

anus --version

docker pull anusai/anus:latest

docker run -it anusai/anus:latest

conda create -n anus python=3.11
conda activate anus

pip install anus-ai

Anus Project Address

• GitHub repository：https://github.com/nikmcfly/ANUS/

Anus application scenarios

• Education and learning: As a learning tool for agent development, it helps beginners understand agent architecture and task planning.
• Prototype development: Provides a basic framework to help developers quickly build agent project prototypes.
• Task automation: Used for simple task planning and execution, such as data processing, file operations, etc.
• Tool integration reference: Provides a reference for projects that need to integrate external tools.
• Open source collaboration: Attracts community developers to participate in improvement and promotes the development of agent technology.