Skip to content
PM Master Handbook ⬇ PDF
Module 17

Replit and Other AI App Builders

Build and host full-stack apps in the browser with Replit Agent, and know when to choose Bolt, v0, Cursor, GitHub Copilot or Figma Make instead.

≈ 5 h studyDraftedSources: Replit docs (docs.replit.com) and pricing; vendor docs for Bolt, v0, Cursor, GitHub Copilot; independent 2026 pricing reviews; checked 23 Oct 2026

On this page
  1. Replit in one picture
  2. Level 1 · Build your first app
  3. Level 2 · Data, secrets and quality
  4. Level 3 · Deploy, schedule and integrate
  5. Labs
  6. The other AI app builders
  7. Common mistakes
  8. Exercises with model answers
  9. Key takeaways
  10. Check your understanding
  11. Recommended resources

Fast-moving tools

App builders change pricing and features monthly. Treat the comparisons as a starting point and check each vendor's pricing page. Screens are illustrative wireframes.

17.1 Replit in one picture

Replit is a complete coding environment in your browser: editor, terminal, database, hosting and an AI Agent that can build a full-stack app (front end, back end and database) from a description. Nothing to install, and it works on a phone too.

17.1.1 What it costs (checked 23 Oct 2026)

Plan Price Notes
Starter Free Daily Agent credits, one published app, limited Agent capability
Core about $20 a month billed yearly (about $25 monthly) Monthly usage credits, more apps, collaborators, custom domains
Pro / Teams about $100 a month For teams with heavier use

Replit uses effort-based pricing: a bigger Agent task costs more than a small one, and usage beyond your plan's credits is billed. Set a usage alert in billing settings. Check replit.com/pricing.

17.2 Level 1 · Build your first app

Level 1 · Beginner

  1. Sign up at replit.com, click Create, and describe the app to Agent (use the five-part prompt from Class 16: what and who, pages, look, data, not now).
  2. Review Agent's plan, remove extras, and approve.
  3. Test in the preview. Ask for one change per message.
  4. If a change goes wrong, roll back to the previous checkpoint.

17.3 Level 2 · Data, secrets and quality

Level 2 · Working PM

  • Database: ask "Store feedback in the database with date, customer segment, text and theme." Use the Database tab to view rows and run SQL such as SELECT theme, COUNT(*) FROM feedback GROUP BY theme;.
  • Secrets: add ANTHROPIC_API_KEY under Secrets. Ask Agent to read it from the environment in server code only.
  • Ask for tests: "Add tests for the form validation and run them."
  • Admin protection: "Protect the admin page with login; never show admin data on public pages."
  • Explain the code: ask Assistant "Which file handles the form? Explain in plain language."

17.4 Level 3 · Deploy, schedule and integrate

Level 3 · Advanced

  • Deployments: choose the type the app needs: static site, autoscale web app (scales with traffic), reserved machine (always on) or scheduled job.
  • Custom domain: Deployments → Settings → link your domain and add the DNS records shown.
  • GitHub: connect Git to push the code to a GitHub repository, then work on it with Claude Code (Class 15).
  • Integrations: Slack incoming webhooks, Stripe in test mode, the Claude API through Secrets.

17.5 Labs

Lab 1 · Feedback hub (Free, 30 minutes)

Build the feedback form and admin table from the picture above with sample data. Roll back once on purpose, to learn checkpoints.

Lab 2 · Database and SQL (20 minutes)

Store submissions in the database. Run a GROUP BY theme query in the Database tab and ask Agent to show the counts on the admin page as a bar chart.

Lab 3 · Weekly digest to Slack (Core or API, 40 minutes)

Create a Slack incoming webhook in your practice workspace. Add it and your Claude key as Secrets. Ask Agent for the scheduled job in the workflow above. Run it manually once and check the Slack message.

17.6 The other AI app builders

ToolBest forBackend and dataFree to startGet the code out
LovablePolished MVPs with login and databaseLovable Cloud or SupabaseYes, 5 credits a dayGitHub sync
ReplitFull-stack apps, scripts, scheduled jobs, hostingBuilt-in PostgreSQLYes, daily creditsGit / GitHub
Bolt.newFast in-browser apps, many frameworksSupabase and othersYes, token-limitedDownload or GitHub
v0 (Vercel)Beautiful UI and Next.js appsVercel integrationsYes, limited creditsDeploy to Vercel or copy code
Figma MakeTurning designs into working prototypesLight; prototype focusStarter plan limitsCopy code
Claude artifactsOne-page tools and clickable conceptsNone (front end only)YesCopy or download
Cursor / WindsurfAI code editors for real codebasesWhatever your code usesFree tiersYour own repository
GitHub CopilotCoding help and agents inside GitHub and editorsWhatever your code usesFree tierYour own repository
Firebase StudioGoogle stack apps with Firebase backendFirebaseYesYour own repository
Claude CodeBuilding and changing real repositoriesWhatever your code usesWith Pro or APIYour own repository
Choosing an AI app builder (checked 23 Oct 2026). For a PM, the usual path is Claude artifacts or Lovable for the first concept, then Claude Code on the GitHub repository once the project gets serious.

17.7 Common mistakes

  • No usage alerts: effort-based pricing can surprise you. Set alerts on day one.
  • Secrets in code: use the Secrets tab or environment variables.
  • Public admin pages: protect anything with customer data behind login.
  • Tool-hopping: pick one builder per project and finish the MVP first.

17.8 Exercises with model answers

Exercise · Pick the builder

(a) A pricing-page concept for a design review in two hours. (b) A nightly script that emails a sales summary. (c) Add a "save for later" feature to your company's existing web app.

Model answer

(a) Claude artifact or Figma Make: fast, no backend needed. (b) Replit scheduled deployment or an n8n workflow (Class 20). (c) Claude Code, Cursor or Copilot on the existing repository, with engineering review, because it changes production code.

17.9 Key takeaways

  • Replit gives you editor, AI Agent, database, secrets and hosting in one browser tab.
  • Checkpoints make experiments safe; usage alerts keep costs safe.
  • Choose a builder by what you are making; prefer tools that export to GitHub.

17.10 Check your understanding

Check your understanding

In Replit, what lets you undo an Agent step, including database changes?

  1. Secrets
  2. Checkpoints and rollback
  3. Deployments
  4. The Shell
Why: Agent saves checkpoints you can roll back to.

Which tool is designed to change an existing production codebase?

  1. Claude artifacts
  2. Figma Make
  3. Claude Code or Cursor
  4. A landing page builder
Why: Coding agents and AI editors work inside your existing repository.