Newly review · Hands-on test · October 25, 2026

Newly Review: We Built and Exported a Native Mobile App

Newly is an AI-native mobile app builder that generates React Native / Expo apps for iOS and Android from natural-language prompts. We used it to build StackCheck, a subscription and software-buying tracker, then pushed it through billing changes, reminders, authentication, cloud data, source export and an Android build.

Disclosure: Newly provided Toolessence with complimentary access (a $25 plan with 100 credits, ending October 30, 2026). This review is not sponsored. Newly had no review, editing or approval rights, and the complimentary access did not determine the verdict. Pricing update, October 28, 2026: Newly's main pricing page and FAQ advertise $25/month with 100 credits/month, also confirmed by the partner in writing. This is no longer merely an unknown test-account allowance. Some secondary official pages still show 50 credits for $25/month, and the credit-usage documentation remains inconsistent — check the applicable terms before buying.

Visit Newly See the Decision Passport

Verdict

Newly was more capable as a development agent than its testing infrastructure was reliable. It generated a real native app, handled iterative changes, built a backend with authentication and per-user isolation, debugged its own backend bug, and produced a meaningful source export and a downloadable APK. The embedded iOS simulator, however, repeatedly failed to launch within its time limit, small UX bugs crept in during iteration, and the export still depends on some Newly-hosted services. We did not assign a numerical score.

At a glance

What it is: AI-native mobile app builder generating React Native / Expo apps for iOS and Android from prompts.
Test date: October 25, 2026
Test project: StackCheck — a native app for tracking software subscriptions and buying decisions.
Model used: Default Newly Agent, unless stated otherwise.
Tested on: Newly's embedded iOS simulator. Android was built but not run on a device or emulator.
Exports: Full source ZIP downloaded and inspected; standalone Android APK generated and downloaded.
Overall conclusion: More capable as a development agent than its testing infrastructure was reliable.

First generation: a working native app

Our first prompt asked for a dashboard, saved software tools with full create/edit/delete, search and filtering, monthly spend calculations, local persistence, and iOS and Android support. Newly generated a working React Native / Expo app. On the iOS simulator we verified the dashboard, the initial $100 monthly spend across three sample tools, the tool detail view, adding and editing a tool, dashboard recalculation, and deletion with a confirmation step.

Newly also added things we did not ask for, such as a yearly-cost presentation, haptics and extra visual behavior. Instruction adherence was good, but not perfect.

StackCheck dashboard in Newly's iOS simulator showing $100 monthly spend across 3 tools with status counts and a category breakdown. — Proves: the generated app ran in Newly's iOS simulator and rendered a dashboard with $100/month across 3 tools.

StackCheck tool detail screen for Notion with status, price, annual equivalent, category, rating, website, notes and a delete control. — Proves: the app included a structured detail view (status, price, annual equivalent, notes, delete), not a static mockup.

StackCheck dashboard after adding Canva: total $115 per month across 4 tools with a new Design category. — Proves: adding Canva moved the total from $100 to $115 and the count from 3 to 4 tools; a Design category appeared.

StackCheck delete confirmation dialog for Canva. — Proves: deletion asked for confirmation before removing the record.

Iteration test: monthly vs annual billing

We asked Newly to add a Monthly / Annual billing frequency without redesigning the app. For an annual $120 tool, StackCheck correctly normalized the cost to $10/month. The dashboard, tool cards and detail view updated, and existing records and features were preserved. This change consumed roughly 3 credits in our account. Newly is investigating this observation; we do not label it a confirmed bug.

Native feature test: renewal reminders

Next we asked for an optional renewal/trial date and a local notification three days before it. Because this required a notifications module, Newly ran fresh native builds — and both the iOS and Android builds succeeded.

On the iOS simulator we verified the date workflow, the reminder toggle, permission behavior during the test, the reminder dates displayed in the app, and the reminder state in the interface. We did not independently verify that the operating system ultimately fired the scheduled notification.

UX issues we observed

The date picker could let taps pass through to controls behind it.
Some reminder helper text did not always restore correctly after changing dates.

Newly agent log stating both native builds succeeded because of the new notifications module, next to the iOS simulator still updating. — Proves: the agent added a native notifications module and both fresh native builds succeeded; the simulator was slow to start during the same workflow.

Authentication and cloud backend

We then asked for authentication and cloud persistence. Newly created and deployed a backend, and during implementation the agent found and fixed a real bug in the import route's authentication middleware.

We verified:

sign-up, logout and login, and session persistence across an app restart
cloud records persisting after login, and local-to-cloud import
create, edit and delete on cloud records, with correct dashboard calculations
a second account initially seeing none of the first account's data
API-level user-isolation tests with two accounts, and unauthenticated requests being rejected

Simulator reliability: useful, but not consistently reliable

Newly's embedded iOS simulator genuinely worked — we used StackCheck interactively in it. But it was unreliable during our test. Multiple launches and updates failed because the remote simulator installation exceeded Newly's 20-minute control-plane deadline; at one point the platform reported four consecutive failures with the same timeout.

The generated app was not shown to be the cause: the iOS and Android builds succeeded, and StackCheck worked whenever the simulator launched successfully. If you need predictable testing infrastructure, plan for this friction.

Newly message reporting that the remote iOS simulator build failed three times in a row because installation exceeded the 20-minute control-plane deadline. — Proves: repeated simulator install timeouts against a 20-minute deadline. This is platform infrastructure, not an app crash.

Models and credit cost

Our account offered several selectable models with different credit multipliers. The default is Newly Agent; alternatives shown included options at 0.05x, 0.15x, 2x and 5x. Credit consumption can therefore depend materially on model choice. Our benchmark used the default Newly Agent throughout, and we have not estimated costs for other models.

Newly currently has no free plan. Its main pricing page at newly.app/pricing and official FAQ advertise an entry plan at $25/month with 100 credits/month, verified October 28, 2026. The partner also confirmed this standard allowance in writing. Our October 25 complimentary test account showed a $25 plan with 100 credits; that historical observation is separate from the current published pricing and later partner confirmation. The 100-credit allowance is no longer merely an unknown test-account figure.

Some secondary official pages, including newly.app/create and newly.app/mobile-app-builder, still show 50 credits for $25/month. The public documentation is therefore not fully consistent. Newly's FAQ describes per-interaction credit pricing, while its Terms describe variable consumption based on the work the agent performs — reading the project, writing code, building, checking and automatic fixes. In our test, individual requests sometimes consumed multiple credits, including roughly 3 credits for the monthly/annual billing iteration. Newly is investigating that observation; it is not a confirmed bug. Check the applicable pricing and credit terms before buying rather than assuming a fixed one-credit-per-prompt cost.

Newly model selector listing Newly Agent as default and alternative models with credit multipliers such as 0.05x, 0.15x, 2x and 5x. — Proves: model choice changes the credit multiplier; Newly Agent is the default.

Source export and lock-in

We downloaded the full project ZIP and inspected it ourselves. It was substantial: React Native / Expo app code, TypeScript and configuration, backend code, database schema and migrations, authentication logic, tests, deployment-related code and assets. The backend used broadly standard technology — Hono, Drizzle, Postgres/Neon-related infrastructure, Cloudflare Workers and Better Auth.

The export still contained Newly-specific dependencies and infrastructure references, including @newly/sdk-react-native, Newly analytics, Newly-hosted API/auth endpoints, Newly build-provenance tooling and platform-injected backend configuration. Analytics were also enabled by default in the tested project's settings.

Source-code exportability

High

Operational portability

Medium

Leaving Newly appears feasible without rewriting the application, but it would require a real infrastructure migration rather than deploying the ZIP unchanged.

Newly project settings for StackCheck showing the analytics toggle enabled and a Download ZIP button for the complete project source. — Proves: a full-source Download ZIP exists (app code, configuration, assets, Newly backend code) and analytics were on in this project.

Android APK

Newly generated a standalone production Android APK: status Ready, 133.8 MB, downloadable directly from Newly, with bundled JavaScript. We downloaded and statically inspected it. APK successfully generated and downloaded; physical-device installation was not tested.

Newly Android build panel showing a production APK with status Ready and a 133.8 MB download. — Proves: a production Android APK reached Ready status at 133.8 MB. It does not prove the app runs on a device.

What we tested vs what we did not verify

Tested hands-on

First-generation app on the iOS simulator: dashboard, detail view, add, edit, delete with confirmation, dashboard recalculation
Monthly / annual billing iteration ($120/year normalized to $10/month)
Renewal/trial date, reminder toggle, permission behavior and displayed reminder dates
Sign-up, logout/login, session persistence across restart, local-to-cloud import, cloud CRUD
Two-account isolation at the UI and API level; unauthenticated requests rejected
Source ZIP download and independent inspection
Android APK generation and download, plus static inspection of the APK

Not verified

That the operating system actually fired the scheduled reminder notification
Android on a physical device or emulator (the build succeeded; the app was not run)
Installing the APK on a phone
Deploying the exported ZIP outside Newly
App Store / Google Play submission
Long-term credit use beyond our session, and non-default models

Where the agent implemented something we did not check ourselves, we say so. Newly's own marketing describes its output as production-ready with full source ownership; we did not test those claims as such, and our findings are limited to this session.

Pros, cons and who it fits

Strengths

Working React Native / Expo app from the first prompt
Iterative changes preserved existing records and features
Generated backend with auth, cloud persistence and per-user isolation
Agent found and fixed a real backend auth bug on its own
Substantial source export: app, backend, schema, migrations, tests
Standalone Android APK generated in-platform

Watch-outs

Embedded iOS simulator repeatedly hit a 20-minute install timeout
Small UX/state bugs appeared during iteration
Added features we did not ask for — adherence good, not perfect
Credit use depends heavily on the selected model
Export still references Newly-hosted endpoints and tooling
No device-level verification of Android or notifications in our test

Likely fit: founders and non-specialist builders who specifically want a native mobile app rather than mainly a web app; people who value source-code access; and MVP or prototype work where iterative AI development is acceptable.

More caution: teams that need predictable simulator and testing infrastructure; teams expecting an immediately infrastructure-independent export; and production-critical apps that need rigorous device QA without developer review.

For the structured pre-purchase view, see the Newly Decision Passport or compare options in the best AI app builders for non-technical founders.

Frequently asked questions

What did Toolessence build with Newly?

A native mobile app called StackCheck for tracking software subscriptions and buying decisions: a dashboard, saved tools with full create/edit/delete, search and filtering, monthly spend calculations and local storage, later extended with annual billing, renewal reminders, authentication and cloud data.

Does Newly generate real native code?

Yes, in our test. Newly generated a React Native / Expo project, and the downloaded source ZIP contained the app code, TypeScript configuration, backend code, database schema and migrations, authentication logic, tests and assets.

Was the embedded iOS simulator reliable?

Not consistently. It genuinely worked and we used StackCheck interactively, but several launches failed because the remote simulator installation exceeded Newly's 20-minute control-plane deadline, including a run of consecutive failures. The iOS and Android builds themselves succeeded, so we treat this as platform infrastructure friction, not a crash in the generated app.

Did the reminder notifications work?

We verified the date workflow, reminder toggle, permission behavior and the reminder dates shown in the app. We did not independently confirm that the operating system fired the scheduled notification. We also saw a date picker that let taps reach controls behind it and helper text that did not always restore after changing dates.

Can you leave Newly without rewriting the app?

It appears feasible. Source-code exportability was high. Operational portability was medium: the export still referenced Newly-specific pieces such as @newly/sdk-react-native, Newly analytics, Newly-hosted API/auth endpoints, build-provenance tooling and platform-injected backend configuration, so leaving would require an infrastructure migration rather than deploying the ZIP unchanged.

Did you test the Android app?

Newly generated a standalone Android APK (status Ready, 133.8 MB) that we downloaded and statically inspected. We did not install it on a physical Android device or run it in an emulator.

How much did the test cost in credits?

The monthly/annual billing change consumed roughly 3 credits in our account using the default Newly Agent. Newly is investigating this observation; we do not label it a confirmed bug. The FAQ describes per-interaction credit pricing, while the Terms describe variable credit consumption based on the work the agent performs. Individual requests sometimes consumed multiple credits in our test. Newly also offers models with different credit multipliers; we did not benchmark those models.

What is Newly's advertised entry plan?

Newly currently has no free plan. Its main pricing page and FAQ advertise $25/month with 100 credits/month, verified October 28, 2026, and the partner confirmed this standard allowance in writing. Our October 25 complimentary test account also showed $25 with 100 credits, but the current allowance is no longer merely an unknown test-account figure. Secondary official pages, including /create and /mobile-app-builder, still show 50 credits for $25/month, and the FAQ's per-interaction description differs from the Terms' work-based credit model. Check the applicable pricing and credit terms before buying.

Was this review sponsored?

No. Newly provided complimentary access (our October 25, 2026 test account showed a $25 plan with 100 credits, ending October 30, 2026). That access did not determine the verdict, and Newly had no editorial control over this review. Separately, the advertised $25/month entry plan with 100 credits/month was verified on the main pricing page and FAQ on October 28, 2026 and confirmed by the partner in writing.