Quickstart: the minimal and easiest setup

Here is the fastest way to get you started - if you want to understand things, have more control / other options please see the detailed installation / configuration documentation. We use Serveo for HTTPS tunneling here and run it in a GPT and assume you are running on a Unix-like system like MacOS, Linux or possibly Windows with Cygwin (not tested). You need a paid ChatGPT account.

  1. Create a directory where the executable should be put and download the jar from the latest release. Rename or symlink the jar to co-developer-gpt-engine.jar.
  2. Create a script codeveloperengine in the same directory with the content from here and make it executable. Symlink it into a directory in your PATH. Alternatively, you can put both files into a directory in your PATH.
  3. Create a directory $HOME/.cgptcodeveloperglobal/ and a file config.properties in there with a content
gptsecret=somewildpassword

where you replace somewildpassword with a long password of your choice. It'll be used to authenticate the GPT requests.

  1. Create an executable file $HOME/.cgptcodeveloperglobal/tunnel.sh with the content
#!/bin/bash
exec ssh -T -R yourenginedomain.serveo.net:80:localhost:3002 serveo.net

where you replace yourenginedomain with a prefix of your choice. It'll be used to access the engine from ChatGPT.

  1. Start codeveloperengine in a directory you want to access with ChatGPT. It will start a webserver on port 3002 and a tunnel to it via serveo.net. You can check whether it works by calling https://yourenginedomain.serveo.net/ in a browser.
  2. Create a GPT according to the documentation.

Now you can use the engine from ChatGPT and enjoy exploring the capabilities of your assistant. :-)