Using AI to write scripts

Hi!

We heavily use AI for writing visual regression tests, but unfortunately not in the TestComplete IDE, because, as mentioned earlier, their IDE has become obsolete. We use Cursor (but it also works with VS Code + Copilot).

TC projects are written in JavaScript, so we can add extensions like ESLint for better code analysis. There's also an unofficial TestComplete extension that provides access to all the engine's specific methods (aqstring, aqobject, etc.), so we can have IntelliSense within the IDE. Git management is also much better. Namemapping hasn't been used for years; we prefer to use Find with a component repository (we use a POM architecture, which helps a lot).

If you have the application's source code available (accessible via Cursor or Copilot), they can easily generate tests that take into account the specifics of the agent.md file.

As a bonus, it even handles component identification automatically because it has access to the component IDs. We don't use much of TestComplete anymore besides the engine (we're still dependent on it because we have applications with WPF/Delphi/Web interoperability -_-' we're not perfect either :-D).

You can also launch TestExecute from VS Code via the command line, but you won't have access to debugging.

I do use VS and Copilot (Copilot is our approved AI tool) to create helper scripts. I've created an MD file to make sure it uses the TC built-in functions and does proper formatting and structuring of the code. 
We use Python for our scripts. While it's usage is limited, it has saved me time when I have been able to use it. Yes, you do have to toggle IDE's but it's easily managed with git. 

A few notes about our setup:

• We don't use namemap. It is too slow and clunky to work with. We have other methods to identify fields. Also, using namemap locks us into TestComplete
• We aim to keep all of our TestComplete specific logic in one class. Everything else is standard python
• TestComplete's IDE is a decade behind other tools and needs a major upgrade. This is one of the main reasons we are looking to use other IDE.

One thing to consider is how TestComplete scripts are typically structured. In many projects, a large portion of the script (often 70–80%) is made up of interactions with NameMapped objects — clicking controls, setting values, checking states, etc. That type of code is very tool-specific and already fairly declarative.

Because of that, using an external IDE with AI assistance can feel like overkill for pure UI interaction, especially when you have to switch back and forth between IDEs and still rely on TestComplete to run and debug the scripts.

Where AI tools may add more value is outside of direct UI actions — for example:

• Helper or utility functions
• Data-driven logic
• String/file handling
• Simplifying or explaining more complex control flow

In those cases, selectively using AI to assist with small, focused pieces of logic may be more practical than trying to author entire TestComplete scripts in an external IDE.

🤖 AI-assisted response
👍 Found it helpful? Click Like
✅ Issue resolved? Click Mark as Solution

Personally, I don't think it's necessary to rely on AI to write automation scripts, as scripting is fairly easy to do on your own. However, when working with complex code that involves concepts like Classes, Polymorphism, Inheritance, Encapsulation, etc., in Object Oriented Programming languages like Python, AI can be extremely helpful for guidance. It saves you from having to think about how to create and pass parameters to objects, but relying on AI too much can also make you forget the fundamentals of programming.

There's no official SmartBear plugin available for Visual Studio Code. Additionally, there are limitations when using Python in TestComplete, so code written in VS Code may not work seamlessly in TestComplete.