Tidepool App Forge
Status: Live.
Your muse
Build a vibe coding platform where a person describes a small app and it generates one with a working chatbot built in. Keep it for its owner. Use only made-up data, like a pretend tide-pool club and a made-up creature called the zubrel snail.
Done. Milestone 1 is done.
npm run verifyexits 0: 10 tests pass, the platform caps check passes, and all 4 requirements pass.Checks ran
Done. Milestone 2 is done.
npm run verifyexits 0: 15 tests pass, 8 of 8 requirements pass, and the platform caps check passes.Checks ran, try 2
Checks passed, try 3

The plan: Tidepool App Forge
Describe a tiny app and get a live app with chatbot.
- Owner opens dashboard and clicks New App
- Owner types a short description of a small app plus what the chatbot should do, using...
- Platform shows a plain language plan with screens, data, and chatbot behavior for review
- Owner confirms or edits the plan in chat
See the full plan
Who it's for
A hobby owner who wants to prototype small fun apps.
How it works
- Owner opens dashboard and clicks New App.
- Owner types a short description of a small app plus what the chatbot should do, using only made-up data.
- Platform shows a plain language plan with screens, data, and chatbot behavior for review.
- Owner confirms or edits the plan in chat.
- Platform builds the app in a sandbox and shows progress.
- Owner opens the live preview and clicks around the generated screens.
- Owner chats with the built-in chatbot about the pretend tide-pool club and zubrel snail.
- Owner asks for fixes and rebuilds until happy.
- Owner clicks Publish to get a live link with its own chatbot token.
- Owner finds the app saved in My Apps for later viewing and editing.
Screens
- Dashboard. Welcome back view with New App button and recent builds.
- New App Brief. Text box to describe the small app and chatbot plus made-up data picks.
- Build Progress. Shows plan, sandbox build steps, logs, and errors with retry.
- App Preview with Chat. Live preview of generated screens side by side with working chatbot.
- My Apps Library. Saved projects with status, live links, rename, rebuild, and delete.
What goes in
- Text: Short description of the small app, 1 to 3 paragraphs, plus chatbot personality and made-up data to include.
- Text (optional): Optional follow-up chat to refine features, text, or chatbot answers before publishing.
What comes out
- Text: Live preview link plus build log and publish status for the generated app.
- Text: Chatbot answers inside the preview about the made-up club and zubrel snail.
- File: Generated app code bundle available for download from the build screen.
Models
Model Used for Why Muse Wonder Turn short descriptions into app plans, Next.js code, and chatbot instructions It reasons over text to design small apps and to draft safe chatbot behavior from made-up data. Limits
- Only the owner can describe, build, and publish apps. Visitors cannot start new builds.
- Published child apps cannot build or publish other apps themselves.
- All content is made up, like the pretend tide-pool club and zubrel snail. No real clubs, animals, or personal data.
- Each build and each child chatbot counts against a small daily model budget, so long builds may stop early.
- Server requests are capped near 4.5 MB. Larger files upload to temporary storage first and are deleted after 24 hours.
From your inspo
The person gave text only with no files. They want an owner-only vibe coding platform that turns a short description into a small app with a built-in chatbot using only made-up data.
Takes
- The job of describing a small app and getting a working app back
- A simple create then preview then publish flow
- Playful sample theme with a pretend club and invented creature
Cost per run: about $0.19 (est.). One run uses muse-wonder-1.0 to plan and draft code: about 12k input tokens at $4 per M equals $0.048 plus about 7k output tokens at $20 per M equals $0.14 for a total of about $0.19.
What you asked for 11 of 11 checked and working
- R1: Owner can type a short description of a small app on a New App screen. Works New App screen shows description box with tide-pool club text and chatbot field.
- R2: System generates a working small app with screens and sample actions from that description. Works Preview shows generated screens like Snail Sighting Log with Run buttons and snail table.
- R3: Every generated app includes a working chatbot the owner can chat with in preview and after publish. Works Chat panels show user question and bot answer about zubrel snail in preview.
- R4: Only the owner can start builds and publish. Visitors can view published demos but cannot create apps. Works Visitor demo states only owner can build and building is turned off here.
- R5: All starter templates and demos use only made-up data such as the pretend tide-pool club and zubrel snail. Works Demos and samples mention only pretend tide-pool club and zubrel snail.
- R6: Owner can move between Dashboard, New App, Build Progress, Preview with Chat, and My Apps from a simple nav. Works Top nav shows Dashboard, New App, Build Progress, Preview with Chat, and My Apps across screens.
- R7: Saved app projects and their publish links stay saved in this browser after a reload. Works My Apps still lists Zubrel Snail Care Game with live link after reload.
- R8: Owner can rename, edit the brief, rebuild, and delete a saved app project. Works My Apps shows Open Rename Delete and renamed title Smoke Renamed Game.
- R9: Dashboard shows an empty state with two made-up examples when no apps exist yet. Works Dashboard empty state shows No apps yet with two made-up sample buttons.
- R10: Build and chat failures show a plain error with a retry button and what to try next. Works Chat failure shows plain message with Retry and what to try next.
- R11: Published apps get their own metered chatbot token and show that they were built with Muse Muse and are not an official Meta product. Works Publish shows own metered token and Built with Muse Muse not official Meta product.
Recorded demo Done. Milestone 3 is done.
npm run verifyexits 0 on the first run: 19 tests pass, 11 of 11 requirements pass, and the platform caps check passes.It's live.