Skip to main content

Coding with AI agents

Actions can be build much faster using coding agents such as Cursor or GitHub copilot. Usually, Actions result in a self contained code without many context switches, which makes them a perfect use case for AI-coding.

Resources​

The instructions file​

Download the Anchorpoint instructions file for LLMs. This contains general instructions and documentation in a single Markdown file that can be referenced within the chat context.

Action examples (optional)​

For better results, you can also use Anchorpoint Actions, which are compiled in a single file and are readable by LLMs. Please note that using the examples and the instructions file will consume more tokens.

Usage​

  1. Make sure that you have an Anchorpoint project open
  2. Open the console for debugging. To access the console click on the "..." button in the top right corner.
  3. Download the instructions file
  4. Open your code editor and add your Anchorpoint project. Your root folder in your code editor has to be the Anchorpoint project root folder.
  5. Add the instructions file to your chat context and write a prompt. If you ask your agent to build a new Anchorpoint Action, it will add it to your existing project in the hidden .ap/actions/ folder.
  6. If you create an Action for the first time, you need to restart Anchorpoint so that the Action will be loaded. Any other modifications of the Python code don’t require a restart.

Playground​

Here are some prompts you could try

  1. Create an action that displays a dialogue box listing all the folders (without subfolders) in the project. Name it 'List Folders'.
  2. Create an action that checks whether the project is an Unreal Engine project and displays a toast notification.
  3. Create an action that displays a progress bar for five seconds. Make this process asynchronous.
  4. Create an action that locks all files in the root folder of a project. Name it 'Lock Root Files'.
  5. Create an action that is registered in the folder context menu and that creates a new text file containing the sentence “Hello World”.
  6. Create an action that checks how many files I have in the project and how big the project size is (excluding the .git folder). Do this process asynchronous and show a progress when doing the calculation. Show your final result in a dialog.