Kiwi Code DocsAgents
Built-in AI agents available in Kiwi Code — each powered by a different model and optimized for different tasks.
Agents
Kiwi Code comes with built-in agents — pre-configured AI models ready to use. Each agent is powered by a different AI provider and optimized for different tasks. You can switch between them at any time.
Switching Agents
In the TUI
Use /autocode-select to pick from the full list:
/autocode-selectOr switch directly by ID:
/use <action_id>In Terminal Mode
kiwi --terminal --action-id <action_id> "Your message"Available Agents
Autocode
The default agent when you launch Kiwi Code. A balanced coding agent with strong general-purpose performance across all tasks.
| Provider | OpenRouter |
| Best for | General coding, debugging, refactoring, code review |
| Switch | /use 6a4be8fbc469a7afb3757403 |
Autocode Claude
Powered by Anthropic's Claude. Excels at careful reasoning, long-context understanding, and following complex instructions precisely.
| Provider | Claude (Anthropic) |
| Best for | Complex refactors, architecture decisions, detailed code review, long files |
| Switch | /use 6a4cd4b6c469a7afb3757406 |
Autocode Claude Deep Thinker
Claude with extended thinking enabled. Takes longer but produces more thorough analysis and higher-quality solutions for hard problems.
| Provider | Claude (Anthropic) |
| Best for | Difficult bugs, complex algorithms, multi-step planning, architecture design |
| Switch | /use 6a4cd4e0bc3d3cfea1172619 |
Autocode OpenAI
Powered by OpenAI's latest model. Fast and capable across a wide range of coding tasks.
| Provider | OpenAI |
| Best for | Fast iteration, API development, scripting, general coding |
| Switch | /use 6a4cd19215f87a77ba650f0c |
Autocode OpenAI v2
OpenAI using the Responses API. A newer integration with improved tool use and streaming.
| Provider | OpenAI (Responses API) |
| Best for | Tool-heavy tasks, structured output, API development |
| Switch | /use 6a4cd313e0e2efd4b2655254 |
Autocode Gemini
Powered by Google's Gemini. Strong at large context windows and working with diverse file types.
| Provider | Google Gemini |
| Best for | Large codebases, multi-file analysis, long context tasks |
| Switch | /use 6a4cd65e0fae4e6de5147a78 |
Autocode Qwen
Powered by Alibaba's Qwen via OpenRouter. A capable open-weight model for coding.
| Provider | Qwen (via OpenRouter) |
| Best for | General coding, multilingual projects, open-source alternative |
| Switch | /use 6a4cd992e0e2efd4b2655256 |
Autocode Grok
Powered by xAI's Grok via OpenRouter. Strong at creative problem-solving and unconventional approaches.
| Provider | Grok (via OpenRouter) |
| Best for | Creative solutions, brainstorming, exploratory coding |
| Switch | /use 6a4cdb3c15f87a77ba650f0d |
Choosing the Right Agent
| Task | Recommended Agent |
|---|---|
| General coding | Autocode (default) |
| Hard bugs, complex logic | Claude Deep Thinker |
| Quick iteration, fast edits | OpenAI or OpenAI v2 |
| Large codebase analysis | Gemini |
| Careful, precise refactoring | Claude |
| Creative problem-solving | Grok |
| Multilingual projects | Qwen |
All agents share the same tools
Every agent can read files, write code, run commands, and search your codebase. The difference is the underlying AI model — each has different strengths in reasoning, speed, and code quality.
Quick Comparison
| Agent | Provider | Speed | Reasoning Depth |
|---|---|---|---|
| Autocode | OpenRouter | Fast | Good |
| Autocode Claude | Anthropic | Medium | High |
| Autocode Claude Deep Thinker | Anthropic | Slow | Very High |
| Autocode OpenAI | OpenAI | Fast | Good |
| Autocode OpenAI v2 | OpenAI | Fast | Good |
| Autocode Gemini | Medium | High | |
| Autocode Qwen | Qwen | Medium | Good |
| Autocode Grok | xAI | Medium | Good |