Quickstart: Command Code

Welcome to Command Code!


In this quickstart guide, we'll walk you through the steps to get up and running with Command Code, the AI-powered coding agent that learns your taste of coding style and preferences using both LLMs and continuous learning with meta neuro-symbolic model taste-1

1

Install Command Code

Use your preferred package manager:

Install Command Code

npm i -g command-code
2

Login to Command Code

Login with CommandCode.ai:

Command Code is available globally as command-code or shorter alias cmd CLI.

Login to Command Code

cmd login
Login Flow
  1. Run cmd login
  2. Open auth page in browser (sign up or log in)
  3. Generate your API key
  4. Paste key in terminal
  5. Select Provider: Command Code
Command Code Authentication Open the authentication page? — copy your API key and paste it here. Yes Copy/Paste your API key in the terminal: ************************************* Select provider Command Code (recommended) You are logged in. API key is stored in ~/.commandcode/auth.json file.

Now you're logged in. Let's try it out!

3

Navigate to your project

Navigate to any project directory and start Command Code cmd:

Navigate to your project directory

cd your-project

Start coding with Command Code

cmd
4

Starting coding with Command Code

You can start by providing a prompt like this:

Prompt

Build a date.js CLI with that tells ISO format of date. Use commander.js and pnpm.
Taste Learning

You'll see that in your projects .commancode/taste/taste.md files and folders where cmd has started learning your coding taste.

In the taste.md file, Command Code will have learned something like this:

.commancode/taste/taste.md

# CLI - Use pnpm as package manager. Confidence: 0.50 - Use commander.js for CLI argument parsing. Confidence: 0.50

Command Code will continuously learn your taste, and auto manage the taste files and folders.

As your taste grows, you'll see Command Code taking your taste out of the main taste.md file into separate files and folders for better organization and reflective context engineering.

Meaning only relevant taste context will be used for your coding sessions. For example, if you are working on a React component, only React-related taste context will be used.

Tip
  • You can press Shift + Tab to switch auto accept edits on.
  • You can press Ctrl + C to exit anytime.

Please join our Discord community to share your feedback and get help. We're still in early days and figuring out the best interface for taste. Taste.md files and folders in your projects allow you full git versioning control. We also have a lot more meta learning on roadmap, help us shape it!


Here are the commands you'll use most:

CommandWhat it does
cmdStart interactive mode
/tasteToggle taste learning on/off
/shareCreate a shareable link for the thread
/unshareStop sharing the thread
/resumeResume a past thread
/memoryManage project memory (persistent context)
/providerSelect AI provider (currently support Command Code only)
/clearClear the thread history
/helpShow all available commands