Skip to content
Docs Community

Quick Start

This guide walks you through your first Multi task - from configuring a provider to getting code generated.

1. Choose a Provider

Section titled “1. Choose a Provider”

Multi supports 30+ AI providers. The fastest ways to get started:

Option A: Claude Code (No API Key)

Section titled “Option A: Claude Code (No API Key)”

If you have Claude Code connected:

  1. Open the Multi panel
  2. Click Settings (gear icon)
  3. Create a new profile and select Claude Code as the provider
  4. Choose a model (e.g., claude-sonnet-4-5)

Multi uses your existing Claude Code credentials automatically - no API key needed.

Option B: Gemini CLI (No API Key)

Section titled “Option B: Gemini CLI (No API Key)”

If Gemini CLI is installed and authenticated:

  1. Open the Multi panel
  2. Click Settings → create a new profile
  3. Select Google Gemini CLI as the provider
  4. Choose a model (e.g., gemini-2.5-pro)

Multi uses your existing Gemini CLI credentials automatically.

Option C: Any API Provider

Section titled “Option C: Any API Provider”

For providers like Anthropic, OpenAI, Google Gemini, OpenRouter, and others:

  1. Open the Multi panel → Settings
  2. Create a new profile
  3. Select your provider (e.g., Anthropic, OpenAI, OpenRouter)
  4. Enter your API key
  5. Select a model

2. Start a Task

Section titled “2. Start a Task”

Open the Multi panel and type a task in natural language:

Create a React component that displays a sortable data table
with pagination and search filtering.

Multi will:

  1. Execute - Read your codebase, write code, run commands
  2. Present changes - Show you diffs for each file modified

3. Review & Approve

Section titled “3. Review & Approve”

Multi asks for your approval before making changes. For each tool call, you can:

  • Approve - Let Multi proceed
  • Approve, and don’t ask again - Let Multi proceed and enable auto-approval for this tool call type
  • Reject - Interrupt this tool call and pause until your next input
  • Reject with reason - Interrupt this tool call and continue using your reason

Auto-Approval

Section titled “Auto-Approval”

For faster workflows, you can enable auto-approval in settings:

  • Auto-approve reads - File reads, glob, grep (safe, read-only)
  • Auto-approve writes - File edits and writes
  • Auto-approve exec - Shell command execution
  • Auto-approve todo - Todo/plan updates

You can also tune safety limits:

  • Max consecutive tool retries - How many retries before Multi asks for help
  • Max step size - Max agentic steps before Multi pauses for confirmation to continue

Next Steps

Section titled “Next Steps”