1
TECHNICAL PROMPT
Tell your AI coding agent what the app should do, who will use it, and which records it needs. Then give it this exact foundation so the app stays portable and its interface and CLI work from the same local data.
COPY THIS TECHNICAL PROMPT
Build a local-first web app that runs from a project folder with Node.js 22+. Use a React/Vite interface served by a local Node server. Store user data in `data/workspace.json` inside that same folder. Have the interface and a CLI read and update data through the local server API, with validation, revision checks to prevent conflicting edits, and atomic writes with backups. Make `npm start` build and launch the app at a localhost URL. Require no cloud account or API.
Add your own screens, fields, and workflows after this prompt. Use the highest effort model, that way the agent will verify the build, data validation, conflict checks and backups, then gives you a finished app in one go.
2
AI AGENT
Open the project folder in an AI coding app that can read and edit local files. The app will open in the built-in AI browser.
01 · UI Updates
Working on the app with AI allows you to request updates as you work, no need for new version releases.
02 · Agentic Workflow
Ask the AI to do any task in the app and it will be able to accomplish it, it's like a built in agent that assists you.
03 · Keywords
Ask the AI to setup specific keywords for specific actions. Like 'Hi' opens the app and gives you a daily summary.
04 · Automation
Set up recurring automations inside Codex or Claude, these tasks will run autonomously on a schedule
3
MULTIPLE USERS
A shared folder can hold one common workspace. A private GitHub repository gives each teammate a separate local copy and a reviewable way to exchange code. Choose the workflow that matches how you need to share data.
OPTION A · SHARED FOLDER
Use one shared workspace
For simple handoffs, keep the project and data in a shared folder, let syncing finish, and have one person edit at a time. For simultaneous use, run one server that everyone connects to with appropriate access controls. Separate local servers writing to the same synced JSON file can conflict.
OPTION B · PRIVATE GITHUB
Give each teammate a local copy
Publish the project code to a private repository with GitHub Desktop. Each teammate clones a local copy and pulls, commits, and pushes changes. GitHub shares committed snapshots, not live app records. Keep `data/workspace.json`, backups, and secrets out of Git unless you deliberately decide to share them.
GITHUB DESKTOP · QUICK SETUP
Set up the private team repo
1. In GitHub Desktop, use File → Add Local Repository for the app folder (or create a repository there).
2. Add `data/workspace.json`, backups, and secrets to `.gitignore` before the first publish.
3. Choose Publish repository, and leave “Keep this code private” selected.
4. On GitHub, open Settings → Collaborators → Add people. Invite teammates; each accepts and clones the repository in Desktop.
5. Fetch and pull before editing. Commit and push afterward. Use a branch and pull request when a code change needs review.
READY TO BUILD
Start with one workflow. Let an AI agent build and test the app, then improve it from real use. The finished app runs locally without a database service, hosting plan, or app subscription. AI tools may have usage limits on the free account.
design review
Take your design to a whole new level with professional UI/UX.
Submit your project and Jad will write the custom prompts to upgrade it.
Submit your design
Code review
AI generated code can look fine and still hide security holes or performance problems. Luckily we have Taz on the team, he’s a software engineer who can review your code, so you know it is production ready before you launch.
Submit your code