Anus is an open-source AI agent framework designed for task automation, enabling users to build agents that execute natural language instructions, collaborate in multi-agent systems, and interact with various digital resources. It targets developers, researchers, and AI enthusiasts seeking a flexible and accessible platform for AI-powered applications.

How It Works

Anus employs a hybrid architecture supporting both single-agent and multi-agent collaboration. Its dynamic task planning breaks down complex requests into manageable steps, with adaptive resource allocation and memory management for context retention. Specialized agent roles (e.g., Researcher, Coder) can be defined, facilitating structured inter-agent communication and consensus mechanisms for collaborative problem-solving.

Quick Start & Requirements

• Installation: pip install anus-ai (recommended), from source, or via Docker (docker pull anusai/anus:latest).
• Prerequisites: Python 3.11+, pip, Git. Optional dependencies for document processing ([documents]), browser automation ([browser]), and code execution ([code]) are available.
• Configuration: Requires API keys (e.g., OpenAI) configured via anus init and .anus/config.yaml.
• Documentation: Documentation Site (Note: URL is illustrative, actual URL may differ).

Highlighted Details

• Supports OpenAI models (GPT-4o) and open-source models (Llama, Mistral).
• Comprehensive tool ecosystem includes web interaction (Playwright), document processing (PDF, Office), secure code execution, and multimodal input handling.
• Features a hybrid agent system allowing seamless switching between single-agent and multi-agent modes.
• Offers both a Command-Line Interface (CLI) and an optional Web Interface for interaction.

Maintenance & Community

• Community-driven design with an emphasis on contributions.
• Community support channels include a Telegram Channel.
• Follows Conventional Commits for PRs and has a Contributor Code of Conduct.

Licensing & Compatibility

• Released under the MIT License, permitting commercial use and closed-source linking.

Limitations & Caveats

• Browser automation requires playwright install. Platform-specific system dependencies may be needed (e.g., windows-curses on Windows, python@3.11 on macOS).