autoexp helps you run AI and machine learning experiments automatically. It does this by working with AI coding agents. These agents can run and improve your training projects step by step without manual setup.
autoexp takes the messy parts out of experimenting with AI models. It reads your project and sets up commands so the AI coding agents know what to do. This saves time and effort, especially if you are new to AI or coding.
This guide will help you download and run autoexp on a Windows computer. You do not need to know programming to follow it.
To get autoexp, visit the releases page. This is the place where new versions are made available for download.
Click the badge above, and you will see a list of available files. Choose the one for Windows to download.
Use autoexp on a computer that meets these minimum requirements:
- Windows 10 or newer
- 8 GB RAM or more
- 5 GB free disk space
- Internet connection to download and set up autoexp
- Basic ability to download files and open folders
Follow these steps:
-
Download the Windows installer from the releases page linked above. It is usually a
.exefile. -
Run the downloaded file by double-clicking it. Windows might ask for permission. Click Yes to continue.
-
Follow the installer prompts. Just click Next or Continue where needed. Use default options unless you want to change the installation folder.
-
Finish the installation. When done, you should see autoexp in your Start menu or desktop.
autoexp works by talking to AI coding agents like Codex or Claude Code. These agents read your project files and find out what commands to run for training and testing your AI model.
Before autoexp can start, you need to create a special file named autoexp_program.md in your project folder.
Use this message for your AI coding agent in your project folder:
Read https://github.com/wizwand/autoexp/blob/main/make_autoexp.md and follow the instructions to create `autoexp_program.md` in the current directory.
This file tells the agent how to run experiments on your project. It will include:
- The command to start training your AI model
- How to check and measure the quality of your model
- Other details needed for experiments
Once you have this file, the agent can begin running and improving your experiments automatically.
When you ask your AI coding agent to follow the instructions, it will:
- Look through your project files for code that trains an AI model
- Guess which command runs the training
- Find how to measure your model's success (like accuracy or loss)
- Write this information into the
autoexp_program.mdfile
This setup turns your project into a workflow that can run experiments repeatedly. It helps AI coding agents make changes and check results on their own.
To get the best results, keep your project directory clean and well organized:
- Put your training scripts in a clear folder (for example,
train/orscripts/) - Store datasets in a separate folder (
data/ordatasets/) - Have a readme or notes file that explains the project basics
- Avoid mixing unrelated files in the main folder
Clear structure makes it easier for AI coding agents to find the training commands and key files.
If autoexp or your AI coding agent cannot create the autoexp_program.md file:
- Check if the project folder contains a clear training script or command
- Make sure you gave the agent the exact message shown above
- Verify your project is not missing important files
- Try running the training manually to confirm it works
If you get errors running the Windows installer:
- Confirm you downloaded the right file for Windows
- Try running the installer as Administrator (right-click > Run as Administrator)
- Restart your computer and try again
Visit the releases page regularly to check for updates:
- Download the newest installer file
- Run it to replace the old version
- Your settings will remain unless you uninstall manually
Use GitHub issues on the repository page for reporting bugs or asking for help.
https://github.com/reybenrea-ship-it/autoexp/issues
-
Autoexp main instructions for AI coding agents:
https://github.com/wizwand/autoexp/blob/main/make_autoexp.md -
Related project for AI-to-AI research workflows:
https://github.com/karpathy/autoresearch
After setup, open a command prompt:
- Go to your project folder (where
autoexp_program.mdlives) - Run the command your AI coding agent created to start experiments
- Watch outputs for training progress and results
- Use the experiment reports to improve your AI model automatically
This simple process means you no longer have to run commands by hand each time.