Interactive Mode

Interactive Mode is the core experience of Kodezi CLI, a conversational, AI powered development assistant that runs directly inside your terminal.

Use natural language to analyze code, generate changes, refactor, run commands, debug issues, and work iteratively with full project context.

Starting Interactive Mode

# Navigate to your project
cd /path/to/project

# Start kodezi-cli
kodezi

Once launched, you’ll see the interactive shell with a prompt ready for your instructions.

The Kodezi CLI interactive mode showing the splash screen and input prompt

Usage

Interactive mode accepts plain natural language, no special syntax:

> "Show me the authentication logic"
> "Add error handling to the login function"
> "Find and fix all TODO comments"
> "Run the tests"
> "Refactor the API service for better structure"

Example of a natural language conversation with the AI assistant

The AI maintains context throughout the conversation and can handle multi-step workflows.

What You Can Ask

Examples of supported tasks:

• Code analysis:
  "Review the user controller for security issues"

• Code generation:
  "Create a new middleware for IP rate limiting"

• Debugging:
  "Why is this function returning undefined?"

• Editing files:
  "Add input validation to registerUser() in user.ts"

• Searching:
  "Find all references to updateAccount"

• Running commands:
  "Run npm test"

• Documentation:
  "Generate a JSDoc comment for this function"

Example Workflow

Below is a realistic multi-step flow illustrating how developers use interactive mode.

> "Analyze the authentication module for security issues"

> "Yes, apply the suggested fixes"

Key Features

Permission System

Before performing any action that changes files or runs commands, Kodezi prompts for approval.

Example prompt:

⚠️  Permission Required

Tool: edit
File: auth.js
Action: Add error handling

Changes:
  + Line 42: try {
  + Line 48: } catch (error) {
  ...

Approve? [y/N]

The permission prompt asking for approval before modifying files

Options:

• y - Approve and execute
• n - Deny and cancel
• v - View full details

This safety mechanism ensures you maintain control over all file modifications and command executions. Review the proposed changes carefully before approving.

Session Persistence

This allows you to:

• Close terminal anytime
• Reopen and continue where you left off
• Resume multi-day work seamlessly
• Maintain project memory

Sessions store references to your conversation, context, and file actions.

Command Line Flags

Interactive mode supports all global flags:

Auto Approve (Use with Caution)

kodezi --yolo
kodezi -y

Enable Debug Logging

kodezi --debug
kodezi -d

Shows underlying tool operations, LSP events, and MCP communication.

Custom Working Directory

kodezi --cwd /path/to/project
kodezi -c /path/to/project

Useful when running tools globally or integrating with external scripts.

Custom Data Directory

kodezi --data-dir /custom/path
kodezi -D /custom/path

See kodezi --help for all flags.

Exiting Interactive Mode

You can exit anytime using:

> exit

Or:

• Ctrl + C
• Ctrl + D

Sessions are automatically saved and can be resumed later.

Related Commands

• kodezi run → For one-off automated tasks
• kodezi auth → Manage authentication
• kodezi logs → View logs, history, and debug output
• Configuration → Customize CLI behavior