Multi-Agent Framework
Z Code adopts a Multi-Agent Framework architecture, integrating multiple industry-leading command-line AI agents to provide a rich selection of AI coding capabilities. You can configure and use these powerful agent services directly within the application — no need to switch between different tools — achieving a truly all-in-one AI-assisted development experience.
How to Choose the Right Agent Framework
When creating a task, you can choose the most suitable agent framework based on task complexity and scenario. Each framework has unique strengths:
| Agent Framework | Provider | Core Strengths | Recommended Models | Best For |
|---|---|---|---|---|
| Claude Code | Anthropic | Strong cross-file reasoning, excels at multi-step autonomous execution on complex projects, clean maintainable code | Sonnet 4.5, Opus 4.6 | Architecture refactoring, legacy system upgrades, complex multi-file changes |
| Gemini CLI | Excels at frontend development, massive context window, native multimodal support, built-in web search | Gemini 3 Pro, Gemini 3 Flash | Frontend page development, image/design-to-code, Google ecosystem projects | |
| Codex | OpenAI | Excels at bug fixing and deep problem solving, outstanding code analysis and diagnostics | GPT-Codex-5.3, GPT-5.2 series | Bug fixing, deep issue investigation, code diagnostics and optimization |
Switch Agent Framework
During a conversation, you can switch between different agent frameworks at any time via the top menu, without needing to create a new task. Each framework has its own strengths — flexible switching lets you get the best AI assistance within the same workflow.
Workflow Suggestions
- Complex Projects — Use Claude Code (Sonnet 4.5 / Opus 4.6) for architecture design, refactoring, and other deep tasks
- Frontend Development — Switch to Gemini CLI (Gemini 3 Pro / Flash) for efficient page development and UI implementation
- Bug Fixing — Use Codex (GPT-Codex-5.3 / GPT-5.2) for precise issue diagnosis and code fixes