I received an email inviting me to try Scram, a new AI web app building platform. The message said it might suit someone experimenting with app builders and described its free mode as similar to Lovable, but without AI credit costs or daily usage limits. That caught my attention.
Rather than use the example prompt they supplied, I wanted to see what it could do with an idea of my own.
A digital home for my Post-it notes
I already use Trello, Notion and Google Drive for planned work. What I wanted was something simpler for the little tasks that pop up during the day: the reminders I would normally write on Post-it notes and stick along the bottom of my monitor. I pictured a private, pastel-coloured board where I could add, move, edit, label, pin and complete notes. This isn’t exactly a new concept but I thought it should be easy to build to test the builder.
Eventually, I’d like to see whether Scram can connect it to ChatGPT through an MCP connector, so I can ask ChatGPT to add a note. That was an idea for later, though. First, I wanted to find out how Scram handled the basic app.
My starting prompt was:
Create a personal daily task post-it style notes dashboard with drag and drop, editable notes. Each note should also be editable and contain the following: date added, label (chip – create, set colour, edit), task description – limited to 500 characters. The dashboard should be private with a single user sign in – no publicly available sign-ups.
Scram asked a few scoping questions before starting the build.
The first scoping questions: board layout, visual style and extra note behaviour
I chose a free-form board, a soft pastel look, and options to pin a note or mark it done so it fades away. Those questions helped narrow the brief without making me write a full specification.
Out of curiosity, I did ask if it could create an MCP connector and the answer was, “Short answer: not as an MCP connection — Scram can’t present an app as an MCP server, which is what ChatGPT’s connector expects. So there’s no way to plug your board into ChatGPT’s connector settings today.
What Scram can do is give your board a private web address that accepts “add a note” requests. Anything that can call a web address could then create a note — but that’s a plain web hook, not an MCP connection, and ChatGPT’s own connector feature won’t accept it.”
Watching the build take shape
One thing I liked was being able to watch the work unfold. Scram showed a project task list and ticked off items such as the data foundation, design system, page structure, access controls and note editing.
Scram’s build plan
It took longer than I’m used to with Lovable, but seeing the stages made it easier to understand what the agent was doing.
When the build was ready, Scram asked whether I wanted it to test the app or open it for me to try. I chose to run it myself.
Scram’s hand-off to the running app, with a suggested checklist for the first test
The setup screen was designed to create the one account for the private board, after which public sign-ups would be closed. The app also warned me that there was no password-reset link, so I made sure to use a password I could keep safe.
First few notes added during testing
The first bugs
The board looked promising, but a couple of basic interactions did not work as I expected. I could see the full 316-character task on the note, yet the edit form opened with an empty text field. The character counter still showed 316 / 500, which made the blank field particularly confusing. I also couldn’t create and apply a label directly as I was adding a note.
The edit form showed the character count, but not the task text
I described both problems to Scram. Its agent inspected the front-end activity and found that the note text had loaded before the edit box existed. In other words, the data was there, but the field had already been created by the time it arrived and never picked it up. That was a useful explanation, and the agent changed the form so it could accept the text whenever it became available.
I also appreciated being able to take a screenshot and paste it straight into the chat. It gave the agent a visual reference for the issue, alongside my description. Something I’ve become used to when interacting with AI chat.
Some fixes, followed by more testing
Scram reported that it had fixed the note text and added a way to create a label inside the note editor. I tried the changes, then found a few more rough edges. The task field had moved below the label controls, while its character count remained near the top of the form. The note text was now editable, but the layout made the relationship between the field and counter hard to follow.
The label colours also needed clearer treatment. A label’s colour changes the note’s background, which I liked, but that wasn’t obvious from the instructions. At one point, the label name and the colour selected for it did not appear to match. I asked Scram to bring the task field back to the top, make the label chip match its colour and explain that the chosen label colour would also colour the note.
The note editor after an update: the task, label and colour controls still needed refinement
Labels were the trickiest part to test and resolve. The interface first showed a row of coloured chips, but it wasn’t immediately clear whether they were labels already attached to the note or choices I could apply. I eventually understood that the chips were toggles: click one to add it to the note and click it again to remove it.
Creating a label introduced another step. I expected that entering a label name and choosing a colour, then clicking Save note, would save both the note and the new label. Instead, I had to use a separate plus button to save the label first. That is a reasonable workflow once you know it, but the button’s purpose wasn’t clear. I asked for it to be labelled “Save Label” and for the label choices to work more like a dropdown, as in Trello.
The label menu after another round of changes, showing the available labels and the separate create-label controls
Even after those changes, I couldn’t complete the label flow reliably. The Save Label button gave no clear indication that anything had happened, and a newly created label did not appear to be available to select in the note editor. I could edit an existing label’s name and colour in the label manager, but I could not confidently apply or remove labels from the note in every test.
One thing I noticed, after the first changes were made is that the app needed to be run again using the Run button (top right), almost as if the preview window wasn’t picking up the changes. This reminded me of some older code compilers. That added another small step to the test loop.
A small surprise in the free mode
While working in the chat, I noticed a context indicator showing how much conversation space was left. I clicked the Compress option to see what it did, and the indicator changed from 78% left to 100% left. There was no clear explanation of what Compress does, how it affects the build, memory or overall project context.
If you’re curious about how these tools work, like me, you’re naturally going to click the buttons and see what happens. I am still unsure why it’s there!
The context indicator before and after using Compress
My first impressions
For a first pass, Scram produced a useable app from a short prompt and a handful of change requests. It asked sensible scoping questions, showed its build progress and explained one of the bugs it found clearly.
The direct screenshot paste into chat was useful, and the absence of build credits in chat mode is an appealing idea.
I spent several rounds clarifying what should happen, then retesting the note editor and label controls.
Some issues were fixed, while others remained unclear or unreliable. That’s where basic UX/UI conventions come into their own: a save button should say what it saves, selected labels should look selected and a control should sit beside the field it describes.
Scram is clearly at an early stage, so I’m not treating this test as a final judgement, but I’m glad I had the chance to try it, and I’ll keep an eye on how it develops.
A low monthly price could be very good value, but features such as MCP connections, Google authentication, email sending, custom domains and higher compute requirements may affect what a finished app costs to run.
For now, I’m sticking with Lovable for my app builds. It has matured into a more intuitive tool, and I can usually give its agents fairly simple instructions without having to return repeatedly to correct basic behaviour. The trade-off is that credits and running costs need to be part of the calculation, especially when you’re building and maintaining several projects simultaneously.
Scram has my attention for now, and compares well to other AI app builders like 8080.AI. I’ll be interested to see how much of this early friction disappears as the platform and its free mode mature.
If you're planning an MVP or early-stage product and want to make sure the foundations are right before you build, we're happy to talk it through.
Start a conversation