Using Retool for large apps / 1000s of tests?

[Peter_Kellner]

First, I want to say that I’m genuinely impressed with what the Retool team has built with the new App Builder.

As a senior React developer and someone who teaches React, I can really appreciate the amount of thought that has gone into the development experience. The guardrails the Retool AI agent provides are valuable, especially compared with just pointing a general-purpose coding agent at a blank repo and hoping it makes good architectural choices.

I also really like that I can have my Claude Code agent look at the Retool-generated source and reason about it, and then effectively have it “talk” with the Retool agent by helping me craft better prompts and review what was generated. That workflow feels powerful.

The area I’m struggling with is testing.

When I build large React apps myself with AI assistance, my normal workflow is that after a feature is built, I ask the agent to also build a regression test for it. That might be Jest, React Testing Library, Puppeteer, Playwright, or whatever fits the feature. The important thing is that once the test exists, I can run it later and know that the app did not accidentally regress.

With the new Retool App Builder, I don’t yet see how that same discipline fits in.

Because the app is primarily built through the Retool AI agent, and because I don’t appear to have the same kind of direct repo/package/test-runner workflow I would have in a normal React project, I’m not clear on where durable automated tests are supposed to live.

For small apps, I can see how manual testing and Retool’s built-in structure may be enough. But I’m building something larger, and my concern is long-term maintainability.

At some point, I expect to turn what I’ve built over to people who may not be senior React developers. I want to make sure they can keep improving the app without accidentally breaking workflows I already built and verified.

So my question for the Retool team is:

What is your thinking around automated testing for larger apps built with the new App Builder?

More specifically:

• Are generated tests on the roadmap?
• Is there a recommended way to add Jest, Playwright, Puppeteer, or similar tests around an App Builder app?
• Should we be thinking of Retool apps as primarily tested through external browser-based tests against a staging URL?
• Is the intended architecture that complex business logic should live outside Retool, in a separately tested service, while Retool owns the UI/workflow layer?
• Are there plans for the Retool AI agent to generate test cases or regression checks as part of feature work?

I’m not asking this as criticism of the product. I’m asking because the App Builder is good enough that I’m now thinking about using it for bigger things. And once an app gets big enough, automated regression testing becomes less of a nice-to-have and more of a core part of how I trust the system over time.

I would love to understand how the Retool team thinks about this, and what best practices you recommend today for teams building larger, longer-lived apps with the new App Builder.

(not sure if this is the right forum to post )

[ChiEn]

Hey, @Peter_Kellner thanks for writing in! We appreciate all the feedback and questions about the new app builder, and absolutely, you are in the right forum to post this! Someone on the product team will be responding here soon!

[Peter_Kellner]

Hi @ChiEn , I'm sure you all are very busy but I'm still hoping to have a discussion or feedback here. I'm about to propose a large project to include retool and I don't have the testing side figure out so that I feel comfortable proposing.

One possibility is to export the project into a stand alone React app and then right my own Pupeteer type E2E tests around every feature. That will mean I have to mock things like airtables and authentication which I think is doable with my own AI skills.md update and merge, but it will be somewhat painful and maybe even a little brittle.

I'd love to hear what you guys are thinking, or event that you are thinkng that for this phase of your development, you will not be expecting people to write large apps that really need this kind of automated testing.

Thanks

[timofey]

Hi Peter, thank you for your patience and for this thoughtful feedback. I’m on the product team and wanted to share a few responses below. I’d also love to learn more about your testing needs, and will reach out directly.

Are generated tests on the roadmap?

We know that regression testing is critical, especially for larger, business critical applications. We want users to build these types of applications with the new app builder, and expect to add a feature for this over time. We don’t have a specific timeline yet, but will share more here as we have updates.

Is there a recommended way to add Jest, Playwright, Puppeteer, or similar tests around an App Builder app?

As you suggested, you can add the tests to your codebase and run them outside of Retool where you have a test runner (e.g., on your local machine, as part of your CI / CD pipeline).

You can get the app code either by downloading it from the homepage app list (click the breadcrumbs near the app and select “Download as ZIP”), or via external source control integration (e.g., GitHub), if you’re on the Enterprise plan.

To run your tests, the tricky part will be mocking out all of the areas that interact with Retool. As you point out, data connectivity through Retool’s resource layer will be the main one (e.g., you’d need to mock all the useYourFunctionName( ) hooks in your frontend code).

“Authentication” I’m a bit confused by, and would love to discuss more in our follow-up chat. We expect all users — builders and end users of applications — to go through the Retool platform authentication layer (vs. having authentication built at the level of each app).

Thanks again for your feedback and hope to chat soon!

[Peter_Kellner]

Sounds like mocking those hooks is the right place. I was worried it would be the airtable access. I've already got a flow where I download the zip file in a local folder and them have enough AI guidance in an external repo that basically inserts the proper stubs so that I can run the app standalone. Regarding auth, I mostly need to test all the different roles (like what anonymous visitors see, what board members see, scholars, etc as well as what they can do. I feel like I'm don't have a clear path to test the backend for proper auth myself (and that worries).

My expectation for auth is I'll use the authenticated user simulator (see screen shot) and have my e2e tests toggle that as first steps. I've not quite figured out the .env solution for how to set things so what I've done is that for my production, I assume all variables that I would assign with env default to the correct values for production, and then let your configuration take over for things like tokens. For my separate project, I can override default .env values as I want to run specific tests.

I love what you guys have done, and since I'm on the board of a non-profit that badly needs a system like this, I plan on building it out for them with the expectation that they can pay for a couple/few hours a month to maintain it when I, for example decide to go to Austria for 3 months on vacation, or, (don't like to think about it), get to old and forgetful to help. I doubt they will find a part time person for what a non-profit can afford that will do things like build tests on a separate stack and repo to verify functionality so I'm really counting on you dev's at retool to plug that gap.