Author
Every AI app builder can run code. Far fewer can run it on a clock that survives your browser closing. Eight builders compared on whether background work runs unattended, where the schedule is actually stored, and whose permissions it runs under.
Every AI app builder can send an email. What separates them is whose domain is in the From field, who your app is permitted to email at all, and how many you get per hour. Eight builders compared from primary vendor documentation.
Push is three jobs, not one: get a device token, store it against a user, and send. Eight AI app builders compared on the middle job, the one that decides whether a notification ever arrives, from primary vendor documentation.
An independent 2026 comparison of what each AI app builder lets you change in a shipped mobile app without a new store review, and where every tool draws the same native boundary.
No AI app builder submits your app to the App Store or Google Play. A primary-source comparison of how far up the build, sign, upload, submit pipeline each one carries you, and the two store gates that no tool can shorten.
All eight AI app builders support custom domains, so the column reads Yes eight times and answers nothing. A primary-source comparison of the three axes that actually decide it: the plan gate, where your DNS terminates, and what freezes when you downgrade.
Every AI app builder can take a payment. The real question is who becomes the legal seller and who takes a cut first. A primary-source comparison of eight builders sorted into four money-path tiers, from own-processor to merchant of record to app-store billing.
MCP support is three unrelated products sharing one name. A primary-source comparison of eight AI app builders on all three: whether an agent can drive the builder, whether the builder reads your tools, and whether the app you ship becomes an MCP server.
Built-in auth is a yes on seven of eight AI app builders in 2026, so it decides nothing. A primary-source comparison on the three questions that do: whose brand is on the consent screen, whose account your user creates, and what the second week costs. Plus the correction that Webflow removed its built-in auth in January 2026.
A 2026 scorecard of vendor lock-in across six AI app builders, scored on five independent layers: code, data, identity, runtime, and what the exit itself costs. Your migration bill is set by your weakest layer, not your average.
A criteria-based 2026 scorecard of database portability across six AI app builders: v0, Bolt.new, Lovable, Replit, Base44, and Bubble. Whose account holds the data, who gives you a connection string, and why a CSV export is not the same as a portable database.
Yes, Lovable is legit: a real registered company, ISO 27001:2022 certified, and its own FAQ grants you code ownership. The real risks are narrower and documented: CVE-2025-48757, effort-based credits, and a Cloud app that can pause at zero balance. Four checks, primary sources only.
