Sthenos. Technologies
Production guide
AI-built apps

Taking a Lovable, Bolt, Cursor, Replit, v0 or Base44 App to Production

An app built with Lovable, Bolt.new, Cursor, Replit, v0 or Base44 is ready for production only after someone has checked what the builder leaves to you: data access rules, secrets, rate limits, logging, backups you have restored, tests, dependencies and spend limits. Lovable and Replit both say their security scans cannot guarantee a secure app, and Base44 says you are responsible for your app's security settings. (Lovable docs: security; Replit docs: security scan levels; Base44 docs: security overview) Use the checks for your builder below, then the fix or rebuild test to decide whether to harden the code you have or replace part of it.

Facts checked against vendor documentation as of November 11, 2026.

Book a Free 30 Minute Consultation→

Want a second pair of eyes before launch? See the production readiness audit.

At a glance

Six AI builders compared

What each builder sets up for you, and the first things to check. Each row links to the full section, where every statement has its source.

BuilderWhat it generatesTop risks to check
LovableTanStack Start (apps from May 13, 2026) or React plus Vite; hosted by Lovable; built-in backend on Supabase's open-source foundation with database, auth and storageRLS policies Lovable calls basic; editor changes hit the live database; daily backups kept about 14 days; out of credits pauses database and auth
Bolt.newJavaScript-only backends; Bolt Cloud hosting, domains and Bolt Database, or your own Supabase on Pro and Teams plansSites without sign-in may get policies letting any visitor change or delete data; full audit on paid plans only; database logs cover one hour
CursorAn AI editor and agent working on a folder of your own code; stack, host, database and auth are yoursAgent edits save to disk at once; .cursorignore does not stop terminal or MCP tools; Workspace Trust off by default; MCP CVEs
ReplitFull stack web app with frontend, backend, API routes and a database; PostgreSQL 16 in development; separate production databaseCritical scan findings block publishing only if a setting is on; rate limiting is yours; usage limits stop services when reached
v0 by VercelNext.js, React, TypeScript, Tailwind CSS and shadcn/ui; deployed to Vercel; databases through the Vercel MarketplacePer-user access is a prompt step in the full-stack guide; projects without GitHub publish straight to production; logs kept 1 hour (Hobby) or 1 day (Pro)
Base44React, Vite and Tailwind CSS; managed NoSQL database, built-in auth and Deno serverless functions"All Users" rules allow access without sign-in; preview writes production data when testing mode is off; backups only on higher plans
Builder-neutral

Checks for every AI-built app

Whichever tool you used, these are the questions to answer with evidence before real users arrive. The builder sections below say what each tool does for you and what it leaves to you.

  • Access control. Every table, entity and server route has a rule that names who may read and who may write, enforced on the server or in the database, and a request made without a session is refused.
  • Secrets. No secret key sits in browser code, a public environment variable or the repository history, and you have rotated one key to prove you can.
  • Rate limits. Sign-in, sign-up, password reset and any endpoint that costs you money per call have a limit you chose.
  • Input validation. Every value that reaches the database or a paid API is checked in server code, not only in the form.
  • Errors. Users see a useful message, and the detail goes to your logs, not to the browser.
  • Logging and alerts. Logs are kept long enough to investigate last week, and an error spike alerts a named person.
  • Backups. You know the backup schedule and retention, and you have restored one into a copy.
  • Tests. Sign-up, sign-in, payment and the core workflow have automated tests that run before each release.
  • Dependencies. The framework is on a version that still gets security patches, and someone owns dependency alerts.
  • Cost limits. Spend limits are set on the builder, the host, the database and every AI or email provider, and you know what happens to the live app when one is reached.
  • Data privacy. You know whether the builder may train on your code and prompts, which plan or setting stops it, and where your users' data is stored.
Builder

Lovable

What it generates

Lovable hosts your app and includes a built-in backend, called Cloud in parts of the product, with a database, storage, authentication, realtime and functions. (Lovable docs: cloud) The built-in backend uses Supabase's open-source foundation. (Lovable docs: cloud)

New apps created from May 13, 2026 use TanStack Start and run server code inside the app; older React plus Vite apps run server code as edge functions. (Lovable docs: edge functions)

You can connect your own Supabase project instead, in which case you manage backups, compute and billing in Supabase. (Lovable docs: supabase) Code can be exported and kept in two-way sync with GitHub, and Lovable says you own your code and data and can host parts or all of your app elsewhere. (Lovable docs: github; Lovable docs: hosting)

What to check before production

  • Auth and row-level security. Lovable sets up row-level security (RLS) policies automatically when it builds features that store user data, but its own guide calls them basic and says to review and adjust them early. (Lovable docs: database access; Lovable docs: security best practices) A policy that checks only for a signed-in user gives every signed-in user the same access to every row it covers. (Lovable docs: database access) Server code that uses the service_role key bypasses your policies, so it must check permissions itself. (Lovable docs: database access; Supabase docs: api keys)
  • Secrets. The publishable database key is visible to anyone using your app by design, so your data is only as safe as your RLS policies. (Lovable docs: database access; Supabase docs: api keys) Values in the Secrets manager are encrypted and never reach the browser. (Lovable docs: secrets) VITE_ values are embedded in the browser bundle at build time and the .env file that holds them is committed to the repository, so nothing secret belongs there. (Lovable docs: secrets)
  • Security scans. A basic scan of database configuration and RLS rules runs automatically on every publish. (Lovable) The Deep scan does not run automatically, and the dependency audit raises only vulnerabilities rated critical. (Lovable docs: security) Lovable says these tools cannot guarantee complete security. (Lovable docs: security)
  • Rate limiting and input validation. Automated abuse protection runs on every published app, and Lovable says Lovable Cloud uses adaptive rate limiting at the IP, user and workspace level. (Lovable docs: hosting; Lovable) In the docs we read on November 11, 2026, we found no per-endpoint rate limit setting, so add limits to sign-in and paid endpoints in your own server code. Lovable's guide says frontend validation provides no security guarantees. (Lovable docs: security best practices) Validate input in server code.
  • Logging and monitoring. Backend log relative time ranges go back up to 5 days, and for long-term retention Lovable suggests building logging or analytics into the app itself. (Lovable docs: logs) Decide who is alerted when errors spike.
  • Backups and rollback. The built-in database is backed up once a day, about 14 days of backups are kept, and you can restore only to one of those daily backups. (Lovable docs: database backups) Editor edits use the same database as your published app, and reverting a version restores code only, not data. (Lovable docs: database; Lovable docs: history)
  • Tests. When you add or change a rule in your app, Lovable adds a small test for it. (Lovable docs: testing) Add your own tests for sign-up, payment and the core workflow.
  • Cost limits. When the monthly Cloud grant and your general credits are both gone, Lovable pauses the database, storage and authentication. (Lovable docs: out of credits) Creating usage limits and alerts needs a paid plan. (Lovable docs: usage limits and alerts)
  • Data privacy. As of September 9, 2026, Lovable may use Free and Pro plan data, including code and prompts, to train its models; Business and Enterprise workspace data are excluded by default, and your apps' end-user data is not used for training. (Lovable docs: data opt out) Opting out does not retract content from datasets or models built before then. (Lovable docs: data opt out)

Incidents on record

CVE-2025-48757, rated 9.3 (critical), records that insufficient RLS policies in Lovable through April 15, 2025 let remote unauthenticated attackers read or write database tables of generated sites. (NVD CVE-2025-48757) Lovable disputes it, on the grounds that each customer is responsible for protecting their own app's data. (NVD CVE-2025-48757)

The researcher reported a March 2025 scan that found inadequate RLS settings on 303 endpoints across 170 of 1,645 projects. (researcher's statement) Lovable 2.0, released April 25, 2025, added a vulnerability check on publish for Supabase-connected projects. (Lovable docs: changelog)

Builder

Bolt.new (StackBlitz)

What it generates

Bolt runs on StackBlitz WebContainers, so the project runs in your browser while you build, and it supports only JavaScript-based backends. (Bolt help center: intro bolt; Bolt help center: supported technologies)

Bolt Cloud covers hosting, domains and databases in one place. (Bolt help center: bolt cloud) Bolt Database uses Supabase technology for its auth email templates and the same checks as Supabase's Security Advisor. (Bolt help center: authentication; Bolt help center: security) Connecting your own Supabase database requires a Pro or Teams plan. (Bolt help center: supabase)

Bolt says the code you create is yours to use for any legal purpose. (Bolt help center: intro bolt) With GitHub sync the Bolt project stays on the main branch. (Bolt help center: git)

What to check before production

  • Auth and row-level security. Bolt turns on RLS for the tables it creates, and a table you add without RLS can be read and changed by anyone. (Bolt help center: security) If you have not asked Bolt to add sign-in, it might write policies that let any visitor view, add, change and delete your data, and sign-in may not be added automatically. (Bolt help center: security; Bolt help center: database)
  • Secrets. Secrets work only with server functions and are never written into browser code or your published site. (Bolt help center: secrets) Values in the project .env file are visible to editors and co-owners, and any value your browser code reads is sent to every visitor. (Bolt help center: secrets)
  • Security scans. The project security audit looks for ways one person could see, change or delete information that is not theirs, and it is available on paid plans only. (Bolt help center: security) You can run up to 30 audits per day. (Bolt help center: security)
  • Rate limiting and input validation. In the Bolt help center pages we read on November 11, 2026, we found none that describes rate limiting or bot protection for Bolt-hosted apps. Put limits and input validation in your server functions.
  • Logging and monitoring. Database logs show only the past hour. (Bolt help center: logs) Built-in hosting analytics come only with paid plans and cannot separate crawler traffic from real visitors. (Bolt help center: analytics)
  • Backups and rollback. Restoring an earlier project version does not change your database. (Bolt help center: rollback backup) In the help center pages we read on November 11, 2026, we found none that describes automatic database backups or point-in-time restore for Bolt Database. Connecting a Supabase database to a project that already has a Bolt database replaces the connection and may cause data loss. (Bolt help center: advanced) A Bolt Database with low usage for six days or more can be paused automatically. (Bolt help center: database)
  • Tests. In the help center pages we read on November 11, 2026, we found no built-in test feature described. Add a test suite before launch.
  • Cost and service limits. Free plan sites stop serving content when they hit monthly data limits and stay offline until the next billing cycle; on Pro, a site pauses when the pay-as-you-go spending cap is reached. (Bolt help center: plans) These limits apply to your Bolt account as a whole. (Bolt help center: plans) The default auth email service is meant for testing and sends at most 2 auth emails per hour for the whole project, so production apps need custom SMTP. (Bolt help center: authentication)
  • Data privacy. Unless you opt out, StackBlitz may use eligible Bolt content for model development and license datasets derived from it to third parties; the opt out is free on every plan. (StackBlitz privacy policy) Enterprise and Team customers under an MSA or DPA are excluded from model training and data sharing. (Bolt)

Incidents on record

An NVD keyword search for "bolt.new" on November 11, 2026 returned no CVE records, and our search on that date found no published Bolt platform security advisory. (services.nvd.nist.gov)

Builder

Cursor

What it generates

Cursor is a desktop AI code editor and agent: you download it, sign in and pick a folder of code to work on. (Cursor docs: quickstart) It works with the source control and tools a team already uses, such as GitHub, GitLab and Bitbucket. (Cursor docs: docs)

Unlike the hosted builders on this page, it does not choose your stack, so your database, sign-in, hosting, rate limits, logging and backups need the builder-neutral checks above and the production readiness checklist linked below.

What to check before production

  • Agent changes. Agents can modify workspace files without approval, except configuration files, and changes save immediately to disk, so Cursor tells you to always use version control. (Cursor docs: security)
  • Secrets. A .cursorignore file hides files from the agent's file tools, but the terminal and MCP server tools used by the agent cannot block access to those files, and Cursor says complete protection is not guaranteed. (Cursor docs: ignore file) Keep production secrets out of the workspace.
  • Agent permissions. Auto-review, the recommended run mode, sends calls it does not already know to be safe to a classifier, and Cursor says the classifier can make mistakes. (Cursor docs: run modes) A Run Everything mode runs every tool call automatically. (Cursor docs: run modes) By default the agent can read files outside the workspace without approval, and Workspace Trust is disabled by default. (Cursor docs: run modes; Cursor docs: security)
  • MCP servers. Cursor warns that MCP servers can access external services and execute code on your behalf. (Cursor docs: mcp) Install only servers you have read.
  • Rules. Rule contents are included at the start of the model context. (Cursor docs: rules) Rules guide the agent; they do not enforce anything in your app.
  • Tests. Cursor's quickstart tells you to review the diff and ask it to run the checks your project already uses: tests, the type checker, linting or a local build. (Cursor docs: quickstart) If the project has none, writing them comes first.
  • Cost limits. On Cursor's billing, leaving the additional usage limit unset means no cap, and enforcement is not instant, so usage can briefly exceed a limit. (Cursor help: spend limits) On legacy Teams plans, on-demand usage is on by default. (Cursor docs: legacy cursor plans) Your app's hosting costs are set by your host, not by Cursor.
  • Data privacy. If Privacy Mode is off, Cursor may store and use codebase data, prompts and code snippets to train its models. (Cursor) Privacy Mode is on by default for team members, and requests still go through Cursor's backend when you use your own API key. (Cursor help: privacy; Cursor)

Incidents on record

CVE-2025-54135: before version 1.3.9, if a .cursor/mcp.json file did not already exist, an indirect prompt injection could make the agent write one and run code without approval; it is fixed in 1.3.9. (NVD CVE-2025-54135; GitHub advisory GHSA-4cxx-hrm3-49rm)

CVE-2025-54136: after a collaborator approved a harmless MCP server, an attacker could swap it for a malicious command without a new prompt; Cursor now asks for approval every time an MCP server entry changes. (NVD CVE-2025-54136; GitHub advisory GHSA-24mc-g4xr-4395)

Cursor's GitHub advisory list shows seven advisories published in 2026 whose titles describe sandbox escapes, including a critical one, CVE-2026-50549, fixed in Cursor 3.0. (GitHub advisory; GitHub advisory GHSA-3v8f-48vw-3mjx) Keep the editor updated.

Builder

Replit

What it generates

Every web app Replit Agent builds includes a frontend and backend, with API routes, a database and server-side logic as the app needs them. (Replit docs: web apps) The development database uses PostgreSQL 16. (Replit docs: development and production)

Development and production databases are separate by default, and the Agent cannot change the production database during development. (Replit blog; Replit docs: development and production) Each checkpoint creates a Git commit. (Replit docs: checkpoints and rollbacks) Replit says you own the code Agent generates and you are responsible for it. (Replit docs: shared responsibility model)

With Replit Auth, your users sign in on a Replit-branded page. (Replit docs: authentication) Apps published from a personal workspace are public by default. (Replit docs: private deployments)

What to check before production

Incidents on record

In a blog post dated July 29, 2025, Replit acknowledged that its Agent deleted data from the database of an app built by a SaaStr co-founder, and said the change reached production because it happened before its new development and production database system. (Replit blog)

SaaStr's own account of the incident says the database held 1,206 executive records and more than 1,196 company profiles. (SaaStr) Replit's stated fix was separate development and production databases by default. (Replit blog) Shared legacy databases were shut down on June 8, 2026. (Replit docs: development and production)

Builder

v0 by Vercel

What it generates

v0 uses Next.js, React, TypeScript, Tailwind CSS and shadcn/ui, and defaults to Next.js for full-stack apps. (v0 docs: faqs; v0 docs: full stack apps)

v0 deploys projects to Vercel: working branches get preview deployments and publishing updates the production deployment. (v0 docs: deployments) You can export the code and deploy elsewhere, but the full experience assumes a GitHub repo connected to a Vercel project. (v0 docs: faqs)

Database integrations such as Supabase, Neon, Upstash and Vercel Blob are added through the Vercel Marketplace, which creates an account on that service and adds environment variables to your project. (v0 docs: databases)

What to check before production

  • Auth and data access. v0's full-stack guide treats per-user access as something you ask for in a prompt, and in the v0 docs we read on November 11, 2026 we found no page that mentions row-level security. (v0 docs: full stack apps) For SQL databases, v0 can generate and execute SQL, including creating and dropping tables. (v0 docs: databases) If you use Supabase, a table without RLS is readable and writable by any role with a grant on it. (Supabase docs: row level security)
  • Secrets. Next.js inlines NEXT_PUBLIC_ variables into the JavaScript bundle at build time, so anything with that prefix is public, and v0 warns about NEXT_PUBLIC_ usage. (Next.js docs: environment variables; v0 docs: security) Vercel's Secret variables are write-only after saving. (Vercel docs: environment variables) In a VM-backed chat, the project's environment variables are available to agent tools in the sandbox, and the sandbox may reach any outbound host by default. (v0 docs: sandbox)
  • Security scans. v0 only runs a Security Review when you ask for one. (v0 docs: security review) In a large project it reviews up to 50 of the riskiest files and skips lockfiles, environment files, credentials and keys. (v0 docs: security review)
  • Rate limiting. Vercel WAF rate limiting is available on all plans, and its counters are tracked per region, so traffic across regions can exceed a single region's limit. (Vercel docs: rate limiting)
  • Logging and monitoring. Vercel keeps runtime logs for 1 hour on Hobby, 1 day on Pro, 3 days on Enterprise and 30 days with Observability Plus. (Vercel docs: runtime)
  • Backups and rollback. Projects without GitHub deploy the chat's current code directly to the Vercel project, and direct code edits do not create a new v0 version. (v0 docs: deployments; v0 docs: versions) In the v0 docs we read on November 11, 2026 we found no page on database backups; they come from your database provider.
  • Tests. v0 can run your project's test command, read the output and fix failures. (v0 docs: terminal commands) It needs tests to exist first.
  • Dependencies. Vercel's LTS policy places Next.js 11.x and 12.x out of support for security patches. (Vercel postmortem) Check the Next.js version in an exported or older project.
  • Cost limits. On Vercel, setting a spend amount does not stop usage on its own, and spend checks run every few minutes, so pausing can lag. (Vercel docs: spend management) v0 itself pauses generation when credits are used up. (v0 docs: pricing)
  • Data privacy. v0 may use prompts and generated content as inputs to third-party models and learning systems to improve its products; Enterprise content and customers who opted out in Vercel settings are excluded. (v0 docs: faqs)

Incidents on record

CVE-2025-29927 let attackers bypass authorization checks done in Next.js middleware; it is fixed in 12.3.5, 13.5.9, 14.2.25 and 15.2.3. (NVD CVE-2025-29927) Next.js deployments hosted on Vercel were automatically protected, while self-hosted apps using next start or standalone output were affected. (GitHub advisory GHSA-f82v-jwr5-mffw; Vercel postmortem) Do not rely on middleware alone for authorization.

Builder

Base44

What it generates

Base44 apps use React with React Router, Vite and Tailwind CSS on the frontend, and a Base44-managed backend: a NoSQL database, built-in user authentication and serverless functions on the Deno runtime. (Base44 docs: introduction)

Base44 lists Mongo for data storage and Render for server services among its subprocessors. (Base44) Two-way GitHub sync requires the Builder plan or higher. (Base44 docs: github)

What to check before production

Incidents on record

Wiz Research reported in July 2025 that a non-secret app_id sent to undocumented registration and email verification endpoints could have let an attacker create a verified account on private Base44 apps, bypassing authentication controls including SSO. (Wiz Research) Wix fixed it within 24 hours of the July 9, 2025 report, and Base44 said it found no evidence that any customer was impacted. (Wiz Research)

Decision test

Fix or rebuild: a seven question test

This test is our method, not a vendor rule. Answer each question with evidence, then read the verdict. It applies to any builder.

QuestionIt passes when
1. Can you take the code with you?The full source is in a Git repository you control, and it builds and runs on a clean machine outside the builder against a copy of your data.
2. Is the data model worth keeping?Every table or entity has an owner or tenant field that access rules can use, and you can write one sentence per table saying who may read and who may write.
3. Can access be enforced on the server?Every route and every table can get a server-side or database rule, and a test that calls each route without a session is refused.
4. Are the secrets clean?A secret scan of the full repository history comes back clean, or every exposed key can be rotated today.
5. Is the stack still patched?The framework and main dependencies are on versions that still receive security patches, and upgrading them does not break the build.
6. Does the platform allow what you need?Nothing your customers or regulators require is a documented platform limit, such as session storage, security headers, backup retention or backend language.
7. Can someone else own it?An engineer who did not build it can list the routes, tables and the rule that guards each one, and can change one flow with the tests still passing.

Verdict

ResultWhat to do
All seven passFix in place: harden, add tests and launch.
Question 4 or 5 fails, the rest passStill a fix: rotate the keys or upgrade, then run the test again.
Question 2 or 3 fails, 1 and 6 passRebuild the data and access layer and keep the interface. This is often the middle path.
Question 1, 6 or 7 failsMove the affected layer off the builder or rebuild it. Fixing code inside a platform you cannot leave or configure does not remove the limit.

Question 6 is where the builder matters most. Examples from the sections above: Base44 stores sign-in tokens in localStorage and does not let you set Content-Security-Policy per app, and Bolt supports only JavaScript-based backends. (Base44 docs: security overview; Bolt help center: supported technologies) Backup retention also depends on the platform and, on Base44, on the plan. (Lovable docs: database backups; Base44 docs: backup and restore; Replit docs: data recovery)

A vendor certification does not answer any of the seven questions. Lovable says its assessment covers Lovable and does not assess your business. (Lovable)

Next step

Score it yourself, or have it audited

Two pages on this site take the checks further.

  • Production readiness checklist. It scores 29 checks across seven areas, from reliability and failure behaviour to ownership and operations, from 0 to 4 on evidence, for a maximum of 116. A single red check blocks the release on its own, and the page publishes no total that means ready to launch. Use it to self-assess.
  • Production readiness audit. Engineers outside your team examine seven areas and deliver a prioritised findings report: what blocks launch, what needs an owner and a date, and what is acceptable risk to carry knowingly, with the evidence behind each finding. Each audit is scoped individually, and our rates are $150 to $250 per hour.

For the security questions alone, see the AI-built prototype security checklist; for the step-by-step work, see from vibe coding to production.

FAQ

Frequently asked questions

How do I take my Lovable app to production?

Review the row-level security policies Lovable generated, because Lovable calls its automatic policies basic and a policy that checks only for a signed-in user gives every signed-in user the same rows. Run the Deep scan, which does not run automatically, and keep secrets out of VITE_ variables, which are embedded in the browser bundle. Plan for editor changes using the live database and for daily backups kept about 14 days. Then work through a production readiness checklist before launch.

What security issues should I check in a Bolt app before launch?

Check that every table has RLS turned on and that a site without sign-in does not have policies letting any visitor change or delete data, which Bolt says can happen. Keep secrets out of browser code and the .env file, and run the project security audit, which is on paid plans. Plan rate limiting and backups yourself, and set up custom SMTP, because the default auth email service sends at most 2 emails per hour.

Is my Cursor-built app production ready?

Cursor does not decide that, because it edits your own codebase and the stack and host are yours. Check the app against a builder-neutral checklist, and check how the agent runs: keep version control on, choose run modes deliberately, keep secrets out of the workspace because .cursorignore does not block the agent's terminal or MCP tools, and update past published advisories such as CVE-2025-54135.

Does Replit keep development and production data separate?

Yes. Replit separates development and production databases by default, and the Agent cannot modify the production database. Production databases have point-in-time recovery with a 7-day window, and restoring the database does not roll back your code.

Are v0 apps secure by default on Vercel?

Not automatically. v0's full-stack guide treats per-user access as a prompt step, its Security Review runs only on request, and NEXT_PUBLIC_ variables are public. Vercel WAF rate limiting is available on all plans, and setting a spend amount does not stop usage on its own. Next.js apps hosted on Vercel were automatically protected against CVE-2025-29927, while self-hosted ones needed the patch.

Can anyone read a Base44 app's data without signing in?

Yes, if a data permission is set to All Users, which Base44 defines as anyone, even without signing in. On a Public app anyone can register, and service-role code in backend functions bypasses row-level and field-level security. Check each entity's rules and run the security scan, which warns but does not block publishing.

Should I fix or rebuild a vibe-coded app?

Fix it when the code is in a repository you control and builds outside the builder, the data model can carry ownership rules, access can be enforced on the server, secrets are clean or rotatable, dependencies are patchable, nothing you need is a platform limit, and an outside engineer can own it. If only the data model or the access layer fails, rebuild that layer and keep the interface. If the code cannot leave the builder or a required control is a platform limit, move that layer off the builder.

Sources

Sources

Every statement about a tool on this page was read in that vendor's own documentation, changelog, pricing or security pages, in the NVD or GitHub advisory record, or in the named researcher's or affected user's own write-up, on November 11, 2026. Vendors change their products often; check the linked page before you rely on a detail. CVE text was read through the NVD API record for the same CVE.

Lovable

Bolt.new

Cursor

Replit

v0 and Vercel

Base44

Supabase (used by Lovable and Bolt)

Published by Sthenos Technologies (Sthenos LLC), McLean, Virginia. Lovable, Bolt, Cursor, Replit, v0, Vercel, Base44 and Supabase are trademarks of their owners; this guide is based on their public documentation.

Launching an app you built with AI?

Score it with the checklist, or ask us to audit it before real users and real money depend on it.

Book a Free 30 Minute Consultation→

Production readiness checklist · Production readiness audit