Prototype to product
Payments, hosting, analytics, GDPR and first users: taking an AI-built prototype live.
From prototype to product: launching an AI-built app properlyTo turn a prototype into a product, get the code into a GitHub repository you own, separate development from production, take payments through Stripe webhooks rather than a success page, add a few analytics events and the UK GDPR basics, then launch to a small group first and fix what they find.Adding Stripe payments to an AI-built app properlyCreate Checkout Sessions on the server, then grant access only from a webhook whose Stripe signature you have verified, using the raw request body. Make fulfilment safe to run twice, test with sandbox keys and test cards, then register a separate live webhook with its own signing secret.Analytics and tracking for a new app, set up rightDecide the three to five questions your MVP must answer, then track only the events that answer them, with one naming pattern. Use a cookieless tool such as Plausible for traffic, send important events like sign-ups from your server, and keep Stripe as the source of truth for revenue. Under UK rules, anything beyond the ICO's analytics exception needs consent.Bolt Stripe payments: setting up checkout that works in productionBolt has a built-in Stripe integration. Add login and a Bolt or Supabase database first, then click your project name, Integrations, Stripe, paste a test secret key and pick your products. Bolt creates the server functions for checkout and webhooks. Test on the published site with Stripe test cards before you swap in a live key.Deploy a Replit app to production: what to check before and after you publishPublish with Autoscale or Reserved VM (Agent apps can't use Static), set every production secret under Adjust settings, review the database migration in the Publish pane, and make sure the server listens on 0.0.0.0 and answers the health check within five seconds. Then test the live URL, not the preview.Deploy a v0 app to production: the publish flow and what to check firstSelect Publish in the v0 chat header: v0 creates or reuses a Vercel project, asks for visibility and a domain, and deploys to a stable production URL. Before you do, set every environment variable for Production in the Vercel project, because the v0 preview only reads Development, and check that GitHub branch rules will let the merge through.Export code from Bolt: what you get, what stays behind and how to run itIn Bolt on desktop, click your project title and choose Download, unzip it, then run npm install and npm run dev. The zip holds your code only. Secrets, database rows, users, uploaded files and deployed server functions stay in Bolt Cloud, so plan how each one moves before you rely on the export.Export code from Replit: getting your app, data and files out in one pieceDownload the code with Download as zip from the file tree menu, or push it to GitHub from the Git pane. Then export the production database separately with pg_dump, copy files out of App Storage and list your secret names. Replit's docs say secrets, the database and deployment settings are not included when you copy a project.Export Lovable code: GitHub sync, zip download and running it outside LovableConnect GitHub, GitLab or Bitbucket from Project settings, then Git. It works on every plan and keeps syncing both ways. Download codebase gives a one-off .zip on paid plans. Neither includes your secrets, database data or stored files, so export those separately and set environment variables before you build.Export v0 code: getting your project out of v0 and running it yourselfThe reliable way to export v0 code is to connect the project to GitHub under Project menu, Settings, GitHub. v0 creates a private repository and pushes the current code, which you can clone. Then pull the Development environment variables with the Vercel CLI, because the keys live in the Vercel project, not in the code.Exporting and owning your AI-built codeIn Bolt, click the project title and choose Download for a zip, or connect GitHub from the GitHub icon for ongoing sync. Either way, the export is only the code. Secrets, data and service settings stay behind, so list your environment variables, export your data and move every account into your own name.GDPR basics for UK founders launching an appBefore launch, check whether you owe the ICO data protection fee (£52 a year for most small companies), record a lawful basis for each use of personal data, publish a plain privacy notice, get consent for non-essential cookies and marketing emails, sign processor terms with your suppliers, and plan for breaches and access requests.Getting your AI-built app ready for real usersBefore real users arrive, check that users can't see each other's data, auth emails actually send, payments are confirmed by webhooks, errors are caught and reported, hosting and database limits won't take you offline, data is backed up, and your privacy notice and cookie setup meet UK rules.How to get your first 100 users after launchYour first 100 users come from you, not from marketing. Pick one narrow group, contact people in it one by one, set them up personally, and ask for feedback rather than likes. Launch posts on Show HN or Product Hunt help only once the product is easy to try.Launch checklist for foundersBefore launch, test every core flow on a real phone, secure your keys and access rules, switch Stripe to live mode with live webhooks, cover the UK GDPR basics, add a sitemap and page titles, set up analytics and error alerts, and plan who answers support in the first week.Lovable Stripe payments: choosing the right setup and making it reliableLovable offers built-in payments (Pro plan or higher, Paddle or Stripe, webhooks managed for you) or your own Stripe account connected with a restricted key. With your own Stripe, webhooks are off by default and subscriptions are checked by email at sign-in, so add a verified webhook and store subscription status before you take real money.Moving off Lovable or Bolt hosting to your own stackSync the code to GitHub and import the repository into Vercel. Newer Lovable apps (TanStack Start) deploy with zero configuration; older Vite apps need a rewrite to index.html. Set every environment variable before the build, add the new domain to your auth redirect URLs, move Stripe webhooks, then switch the domain.Prototype vs MVP vs product: what's the differenceA prototype answers whether an idea makes sense to people. An MVP answers whether real people will use it or pay for it, so it has to work reliably for strangers. A product is what you run once the answer is yes: it needs support, monitoring, backups and room to change.Replit Stripe payments: from sandbox to live without breaking checkoutReplit gives you two ways to add Stripe: the Replit Stripe connector, which starts in a sandbox and goes live through the Replit Integrated Payments app, or your own keys in Secrets. Use one method per environment, test with Stripe's test cards, and only mark an order paid when a verified payment event arrives.Scaling an MVP: what breaks firstIn most MVPs the database breaks first: missing indexes and too many connections from serverless functions. Next come long tasks that time out inside web requests, payment webhooks that run twice or out of order, and errors nobody sees. Fix those in that order before you think about bigger servers.Turning a no-code app into a coded productYou can't export a Bubble app as code, so migrating means rebuilding the app in a coded stack and moving the data across. Export your data with Bubble's CSV or JSON export or its Data API, map each data type and workflow to a table and server function, then switch users over in one planned cutover.v0 Stripe payments: from the sandbox to taking real moneyv0 adds Stripe through the Vercel Marketplace, which creates a Stripe sandbox and adds the API keys to your Vercel project. To take real payments you connect a live Stripe account, redeploy, and add a verified webhook so orders are only fulfilled when Stripe confirms the payment.Writing a product spec for an AI agent or developerA good product spec says who the feature is for, what is in and out of scope, the data it touches, each flow step by step, every state including empty and error, the rules and edge cases, and testable acceptance criteria. The same spec works for an AI coding agent and a developer, as long as it is specific.
See if BlackFig fits your idea (it probably does)
A quick 30 minute call. No slides, no hard sell. You leave with a clear next step either way.