Why Cody Is A Great AI Coding Assistant

Introduction

In this article I want to introduce you to Cody. A colleague showed it to me and I gave it a try for about a week. Here is why Cody free is a great AI coding assistant.

What is Cody?

Cody is an AI extension for code editors and IDEs. It helps developers find bugs, write unit tests, autocomplete code blocks, and more. Cody can be used 100% free and works with various LLMs. The source code is published on GitHub.

This is how the AI describes itself:

Cody is a helpful AI coding assistant created by Sourcegraph! I help developers write, understand, and fix code more easily. Think of me as your friendly coding partner who can assist with tasks like:

Writing new code

Explaining how code works

Finding and fixing bugs

Answering programming questions

Converting code between languages

Suggesting code improvements

I aim to make coding faster and more enjoyable by being your knowledgeable companion throughout the development process. I can work with many programming languages and understand software development concepts to provide practical help when you need it.

Although their website says you only get 200 chat messages per month, I am still at 0/200 after many days of extensive use. Is it a bug? I am not sure but I won’t complain.

How to use it

You can get Cody for VS Code, many JetBrains IDEs, and as a web interface. More options are marked with experimental (Visual Studio, Cody CLI) or coming soon (Emacs, Eclipse).

After the installation in VS Code, Cody offers a chat window as well as an autocomplete feature. The chat window looks like this:

Cody chat window in Visual Studio Code (free version)

There is a chat window with referenced files for context information, a model selection menu, a prompt selection menu, and quick links to all default prompts that come with Cody.

Apart from that, it works like any other AI interface these days. Type in your prompt, give some context, choose a model, and wait for the answer.

In addition, you get an autocomplete feature. While typing, Cody will suggest text to complete your input. The proposals are not always perfect but they will save you time!

This autocomplete feature alone is worth the installation of Cody in my opinion if you don’t already have anything similar in place.

Models

Cody doesn’t have its own LLM but rather uses existing ones. You can choose per prompt which one you want to use. Some models are only available for Pro and Enterprise customers. But as a free user you currently have access to the following models:

Gemini 2.0 Flash
Gemini 1.5 Pro
Gemini 1.5 Flash
Claude 3.5 Sonnet
Claude 3.5 Haiku
Claude 3 Haiku
Mixtral 8x7B
GPT-4o-mini

All supported Cody models of the Visual Studio Code extension in the free version

I am not familiar with all of them but they should all be suited as an every day software developer companion.

Context

You can add files or URLs in the chat window which will then be used to better answer your question about the project. By default, the context always contains your current repository and the active file. If you don’t want that, just delete the entries.

To add new context, type @ in the chat window and choose one of the options that appear in the list.

Cody context import options. After typing @ in the chat window, a list of import options appears.

The dynamic context is one of the reasons why I think why cody is a great AI coding assistant.

Predefined prompts

Cody comes with few predefined prompts that are useful for software developers. Of course, you can ask your own question anytime you want. In the History tab you can find all your previous conversations.

The existing and self-explaining prompts are document code, explain code, find code smells, and generate unit tests. Quite useful but you’ll probably end up writing more prompts for your projects.

Cody prompt window with all public and private prompts. You can create new reusable private prompts from the cody website.

To write a private reusable prompt for your account, you need to go to the website. The interface looks like this and is self-explaining:

Create a new prompt for your account on the Cody website

The created prompt is available immediately in your library, code editors, and IDEs.

Data privacy

When something is free, you usually pay with your data. The privacy policy of Cody however states that

Neither Cody nor their partners train their models on your code without your permission
Context information is sent to LLM but deleted as soon as the output is generated

This is good news!

I couldn’t find a settings to check whether I agreed to model training on my data or not. The only thing I found is a telemetry setting which is enabled by default. Maybe this includes the training consent.

Cody telemetry settings in VS Code. It is enabled by default.

You can also bring your own model and use Cody with it. This could be useful in enterprise environments with stricter data protection regulations.

I would avoid using it with critical or security-sensitive data though.

Is it worth it?

Yes!

✅ Free

✅ Unlimited autocompletes

✅ Unlimited chat conversations (probably)

✅ No LLM training on your data without your permission

⚠️ Chat messages and responses are tracked

⚠️ Privacy policy could be a bit clearer

In my opinion they are really few reasons to go for a paid subscription. You get a better SLA and access to more powerful models.

Conclusion

In this article I showed you why Cody free is a great AI coding assistant. A good chat interface with configurable context information and access to 8 different LLMs result in a great user experience. Users also benefit from unlimited autocompletes and conversations, a private prompt library, and a customer-friendly privacy policy. It looks like a great product overall!

What is your impression about Cody? Is it better than GitHub Copilot Free?