Base44 bug fixing: seven common problems in a Base44 app, and the fix for each
Short answer: most Base44 bugs are one of two things. Either a recent change broke
something, and going back to the version that worked is faster than asking the AI to repair it. Or a setting
outside the chat is wrong: the app was not published again, a data permission, the app's visibility, or a secret.
Find your symptom below; each one has a fix you can do yourself.
Where to look first (two minutes)
Four places tell you what kind of problem it is:
The chat. If an Issues Found message shows in the editor, press Resolve with AI first.
Base44's documentation says this one does not use your credits.
Version history (the icon above the chat): did it work in an earlier version? Then it is problem 1.
Activity Monitor (the More actions menu at the top right of the editor): every request the app makes
in the preview, with its status code and the error that came back.
Dashboard, Monitoring: errors from your live app. And
status.base44.com: if Base44 itself has an incident,
wait before you change anything.
Match the symptom
What you see
Problem
A blank or white screen, or a page that never stops loading, right after a change
A change broke the app. The usual story: it worked, you asked for one more feature, and now a
page is blank. The cause is in the last change, so the fastest repair is to go back to before it.
Fix: hover over the chat message that made the change and press the Revert icon under
it. The app goes back to how it was before that message, and later changes are undone too. To go back further,
open Version history, preview the versions, and restore the last one that worked. Do not type "undo that"
into the chat: Base44's documentation warns that asking the AI to undo makes a new change, so the app may not end
up as it was. If the screen goes blank after saving data, ask the AI to check that the field types of your data
match what the app saves.
The AI is going in circles. Each new "please fix it" message is another guess on top of the
last guess, and each one in Build mode costs message credits. After two failed tries, a third worded the same way
rarely helps.
Fix: stop and revert to the last version that worked (problem 1). Switch the chat to
Discuss mode, which changes no code until you approve it and, by Base44's documentation, uses about 0.3
message credits a message. Describe the bug the way a tester would: which page, what you clicked just before,
what you expected, what happened, and the error from the Activity Monitor copied word for word. Ask for the
cause first and the change second. If the chat says the project is too large, the request was more than the AI
could read at once: ask about one page or one function at a time. When the same bug survives all of that, it
needs a person reading the code.
The live app was not updated. In Base44 the editor is a draft. The live app only changes when
you press Publish again. Two things are easy to miss: a backend function keeps running the version from
your last publish, and a changed secret may keep its old value in the live app.
Fix: press Publish after the change, then reload the live app in a private window so
an old copy in your browser does not fool you. If you edited a backend function, press Save & Redeploy
in Dashboard, Code, and publish again. If you changed a secret, publish again too.
Data permissions are wrong. Every data table has rules for who may create, read, update and
delete. All Users means anyone, even without signing in. Creator Only means a person reaches only
the records they made. User Property Check ties an action to something on the account, such as the Admin
role. Too tight, and users see empty lists or cannot save. Too loose, and every user can read or delete
everybody's records.
Fix: open the table under Data in the dashboard. If a Permission risks detected
banner is there, open More details, read the recommended rules, and apply them. For data that belongs to each
person (orders, notes, messages) use Creator Only on read, update and delete. A mistake Base44's documentation
names: typing id as the entity field in a rule. That is the record's own id, it never matches a
person, and the list stays empty; use a field that holds the person's email, or Creator Only. Then test as an
ordinary user with a second account, not as yourself. Our
security page explains the two-account test.
Visibility, invites or the sign-up code. Who can open the app is one setting (App Visibility:
Public, Private or Workspace) and how they sign in is another (the Authentication page). A Private app lets in
only people you invited, and Base44 says it needs the Starter plan or higher. An invited person must sign up with
the exact email address you invited; another address only sends you an access request.
Fix: check App Visibility on the dashboard's Overview first. For "my code does not work":
the emailed code is good for 10 minutes and 5 tries, and a newer code replaces the older one, so the person
should ask for a fresh code and use only that one. Someone who closed the verification screen before typing the
code has to sign up again with the same address. If a whole office or classroom signs up at the same moment, some
get a "too many requests" error because Base44 limits sign-ups from one internet connection: they wait a minute
and try again. To let people skip the code, offer Google login. With the default Google setup the Google window
says base44.com; your own name there needs your own Google client id and secret on the Authentication
page.
A secret or a backend function. Keys for outside services (an email sender, an AI provider,
Stripe) live on the Secrets page, and backend functions find each one by its exact name. A function that asks
for RESEND_API_KEY does not find Resend_Key. A key pasted into a page's code instead is
visible to every visitor.
Fix: open the Activity Monitor, repeat the action, and read the failing request's Response
tab: it shows the error that came back. Compare the secret's name with the name the function reads, letter for
letter. After changing a secret or a function, publish again (problem 3). Run Security scan from the
dashboard: Base44 says it is on every plan, the free one included, and it lists keys that visitors could reach.
If calls from one function to another fail with a 403 that mentions "a prohibited IP address", Base44's
documentation says the functions are calling each other too many levels deep; retrying does not help, the chain
has to be made shorter.
Something outside your app is spending your credits. Base44 has two kinds of credits. Message
credits pay for your chat with the builder. Integration credits are used by the running app each time it uses
Base44's built-in services: an email, an image, an AI answer, an in-app agent, a workflow step, a backend
function run. If the AI or email feature can be called from outside your app without signing in, a stranger or a
bot can run it and spend them.
Fix: run the Security scan and fix every finding about backend functions that strangers can
call and credit-using features that can be reached from outside. Make those features require a signed-in user.
Look at each workflow's run history for one that runs far more often than you meant. A service you call with your
own API key does not use integration credits for that service, though the backend function run itself still
counts a little.
Test it like a stranger before real users arrive
On the published app, not the editor: (1) open it in a private window,
signed out: you must see only what a visitor should see. (2) Sign up as a brand-new user with a second email
address and do the main task from start to finish. (3) As that second user, look for the first user's records:
you must not find them. (4) Change something in the editor, publish, and check the live app shows it. If all four
hold, the seven problems above are not in your app.
This is a checklist, not a study: we have not counted how often each problem happens. The facts
about Base44 (Revert and Version history, Discuss mode and its cost, publishing, the permission rules, visibility
and plans, sign-up codes, secrets, the security scan, the two kinds of credits) were read in
Base44's own documentation on November 20, 2026; the advice
on describing a bug and testing as a second user is general engineering practice. Base44 changes its menus, plans
and prices often, so check its documentation if a name here does not match what you see. Urgisoft is not
affiliated with Base44.
When the same bug survives the steps above, it needs a person reading
the code. We fix apps made with Base44, Lovable, Bolt, Replit, Cursor and other AI builders, and tell you in plain
words what was wrong. Fixed prices on Fiverr: one bug from $40, up to five bugs with the app working live from
$150. No fix, no charge. For a Base44 app we work through its GitHub connection (Base44 offers that on the Builder
plan or higher) or as a collaborator you invite to the app's editor; we never ask for your password.