> ## Documentation Index
> Fetch the complete documentation index at: https://docs.puzld.cc/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

> Get up and running with PuzldAI in minutes

## Install

```bash theme={null}
npm install -g puzldai
```

Or try without installing:

```bash theme={null}
npx puzldai
```

## Prerequisites

PuzldAI orchestrates existing CLI tools. Install the ones you want to use:

| Agent | CLI | Install |
| - | - | - |
| Claude | [Claude CLI](https://docs.anthropic.com) | Follow Anthropic docs |
| Gemini | [Gemini CLI](https://ai.google.dev) | Follow Google docs |
| Codex | [Codex CLI](https://openai.com) | Follow OpenAI docs |
| Ollama | [Ollama](https://ollama.ai) | `curl -fsSL https://ollama.ai/install.sh \| sh` |
| Mistral | [Vibe CLI](https://github.com/mistralai/vibe) | Follow Mistral docs |

> [!TIP]
>
> You only need ONE CLI installed to start. Claude and Ollama have the best agentic mode support.

## Check Available Agents

```bash theme={null}
puzldai check
```

This shows which CLIs are installed and ready.

## First Run

### Interactive TUI

```bash theme={null}
puzldai
```

On first launch in a new directory, you'll see a trust prompt. This is a security gate (like Claude Code) asking if you trust PuzldAI to read, edit, and execute files in that workspace.

### Single Task

```bash theme={null}
puzldai run "explain recursion"
```

### Compare Agents

```bash theme={null}
puzldai compare "best error handling practices" -a claude,gemini
```

## Basic Commands

> [card] Chat
>
> Just type your message in the TUI. Auto-routing picks the best agent.

> [card] Compare
>
> ```
> /compare claude,gemini "your task"
> ```

> [card] Switch Agent
>
> ```
> /agent claude
> ```

> [card] Help
>
> ```
> /help
> ```

## Keyboard Shortcuts

| Shortcut | Action |
| - | - |
| `Enter` | Send message |
| `Escape` | Go back / Hide view |
| `Ctrl+E` | Expand last result |
| `Ctrl+C` | Cancel operation |
| `Tab` | Autocomplete command |
| `↑/↓` | Navigate history |

## Next Steps

> [card] [Workflows & Automation](https://docs.puzld.cc/modes/pipeline.md)
>
> Pipeline & Workflows, Autopilot mode

> [card] [Multi-Agent Modes](https://docs.puzld.cc/modes/overview.md)
>
> Compare, Debate, Correction, Consensus

> [card] [Agentic Mode](https://docs.puzld.cc/features/agentic.md)
>
> LLMs explore your codebase and propose edits

> [card] [Configuration](https://docs.puzld.cc/configuration.md)
>
> Models, routing, and settings


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.