Mini-Kode is an educational AI coding assistant CLI designed to demystify modern AI coding agent architectures for developers. It offers a complete, yet manageable, 14K-line implementation that balances complexity and simplicity, enabling users to understand, experiment with, and extend AI coding assistant systems.

How It Works

The project employs a clean, education-first architecture built around the OpenAI SDK, featuring thoughtful type design and comprehensive comments. It provides a unified tool system for file operations, search, and command execution, with MCP integration for extended functionality. Mini-Kode supports multiple LLMs (OpenAI, GLM, DeepSeek, compatible APIs) and features a complete agent loop with streaming responses and human-in-the-loop approval, all accessible via a modern React/Ink-based terminal UI. Project context is managed via an AGENTS.md file, offering automatic context awareness and persistence across sessions.

Quick Start & Requirements

• Install: npm install -g mini-kode
• API Keys: Set DEEPSEEK_API_KEY, GLM_API_KEY, or OPENAI_API_KEY. For OpenAI-compatible APIs, set MINIKODE_API_KEY, MINIKODE_BASE_URL, and MINIKODE_MODEL.
• Run: mini-kode for interactive UI mode, or mini-kode "your command" for non-interactive execution.
• Development Prerequisites: Bun, pnpm.
• Development Install: pnpm install.
• Docs: Refer to docs/README.md for detailed system documentation.

Highlighted Details

• A complete, yet manageable, 14K-line implementation focused on educational value.
• Unified tool system supporting file operations, search, and command execution.
• Multi-LLM compatibility, including OpenAI, GLM, and DeepSeek APIs.
• Modern, themeable terminal UI built with React and Ink.
• Project memory via AGENTS.md for context-aware assistance.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), sponsorships, or roadmaps were provided in the README.

Licensing & Compatibility

The license type and any compatibility notes for commercial use or closed-source linking are not specified in the provided README.

Limitations & Caveats

The project is explicitly educational, suggesting it may not be production-ready for all scenarios. Several advanced features are listed under "Future Plans," indicating current limitations, including the absence of session persistence, built-in LLM tool integration (specifically noting DeepSeek's lack of web search tools), image input support, sub-agents, advanced reasoning model support, tool call message streaming, and context caching.