A minimal yet professional demo project, Mini Agent showcases the core execution pipeline and production-grade features for building AI agents with the MiniMax M2 model. It targets developers and researchers seeking a robust foundation for complex, long-running tasks, offering an Anthropic-compatible API for interleaved thinking and a suite of integrated tools. The project simplifies agent development with a clean design and comprehensive features, enabling efficient prototyping and deployment.

How It Works

The agent leverages the MiniMax M2 model through an Anthropic-compatible API, enabling interleaved thinking for advanced reasoning. Its architecture includes a full agent execution loop, persistent memory via a Session Note Tool, and intelligent context management that automatically summarizes conversation history to handle tasks of virtually unlimited length. It integrates 15 professional Claude Skills and MCP tools for web search and knowledge graph access, facilitating complex workflows and debugging.

Quick Start & Requirements

• Prerequisites: uv package manager.
• Installation:
  • Quick Start Mode: uv tool install git+https://github.com/MiniMax-AI/Mini-Agent.git, followed by running a setup script (setup-config.sh or setup-config.ps1) to configure API keys and base URLs in ~/.mini-agent/config/config.yaml.
  • Development Mode: Clone the repository, sync dependencies with uv sync, and edit the config.yaml file.
• API Key: Obtain from MiniMax global (https://platform.minimax.io) or China (https://platform.minimaxi.com) platforms.
• Running: Execute mini-agent or mini-agent --workspace /path/to/your/project.
• Links: Setup scripts available via direct URLs on GitHub.

Highlighted Details

• Full Agent Execution Loop: Provides a complete foundation with file system and shell tools.
• Persistent Memory: Session Note Tool retains key information across sessions.
• Intelligent Context Management: Automatic conversation summarization enables handling long tasks.
• Claude Skills Integration: Includes 15 professional skills for various domains like documents, design, and testing.
• MCP Tool Integration: Natively supports MCP for web search and knowledge graph access.
• Comprehensive Logging: Detailed logs aid debugging.

Maintenance & Community

The project welcomes issues and pull requests, with a dedicated contributing guide and code of conduct available, indicating active community engagement potential.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

A workaround for SSL certificate verification errors is provided for testing, suggesting potential environment-specific configuration needs for secure communication.