Cla - Claude In The Shell
About
A simple CLI chat tool designed for easy interaction with Anthropic’s LLM models. Cla is based on the Cha project.
Backstory
I developed Cha to easily interact with the best LLM model in the terminal and it completely changed the way I interact with my shell. This shows because the project got over 30 stars in the months since I released it and marketed it on Twitter/X and Reddit. But, on June 21, 2024, Anthropic released their latest advanced model, Claude 3.5 Sonnet. After trying out this model through Anthropic’s Pro plan, I immediately saw that Claude 3.5 was far better at coding than OpenAI’s GPT-4 model. I figured this out because Claude 3.5 successfully allowed me to build a complete and complex React app, which was something GPT-4 could not do even with the release of GPT-4o, which I would say was an even worse model compared to the original GPT-4 model. Knowing this, I decided to build Cla to act like Cha but for Anthropic’s models since Claude 3.5 was far better at coding than GPT-4.
Demos
Difference Between Cha
Cla is heavily based off of my last project, Cha. But, Cla uses Anthropic’s models rather than OpenAI’s models. Also, Cla does not support Answer Search, URL page scraping, YouTube scraping, and image generation. So it only supports a basic CLI chat interface, multi-line support, and non-interactive chat mode (through the CLI or file).
Also, you will have to grab an API key from Anthropic. To do this, you can follow Anthropic’s tutorial and/or check out Anthropic’s API Key Page.
Marketing
Currently, I’ve only done marketing of this project on Twitter/X through my main personal account and through Reddit. I had the best success on Reddit, especially through the Anthropic Subreddit. Through this marketing, I got 3 GitHub stars in one week. I released Cla back at the very start of July 2024, I think around July 1st or 2nd.
Conclusion
It was a surprise to me how Anthropic’s Claude-3.5 model surpassed OpenAI’s GPT-4 model, especially when it comes to coding. Due to this, I built Cla to act like Cha but for Anthropic’s models. Since doing this, I found myself using Cla 50% of the time and Cha the other 50% of the time. Overall, I’m excited to see how this project goes.