Field Service Job Costing and Profitability Tracker

AI Prompts to Build This

Copy these build prompts into Claude, Cursor, or your AI coding tool. Create a free account to unlock the full research behind them.

1. 1. Project Setup

   Create a monorepo with a Next.js 14 (App Router, TypeScript, Tailwind) web dashboard and a React Native + Expo mobile app for a field service job costing tool called "MarginTrack."
   
   Provision Supabase with these tables:
   - companies (id, name, trade_type, plan)
   - technicians (id, company_id, name, phone, loaded_hourly_rate_cents, role)
   - jobs (id, company_id, job_type, crew_id, quoted_price_cents, status, opened_at, closed_at)
   - time_entries (id, job_id, technician_id, started_at, ended_at, source ENUM['app','manual'])
   - materials (id, job_id, receipt_image_url, extracted_line_items JSONB, total_cost_cents, added_by)
   - drive_segments (id, job_id, technician_id, started_at, ended_at, miles, source ENUM['gps','manual'])
   
   Enable row-level security so every table is scoped to company_id via the authenticated user's company membership. Set up Stripe Billing with three products: Free, Crew ($49/mo), Company ($129/mo). Add env vars for SUPABASE_URL, SUPABASE_ANON_KEY, ANTHROPIC_API_KEY, VERYFI_CLIENT_ID, VERYFI_API_KEY, QUICKBOOKS_CLIENT_ID.

2. 2. Mobile Time + Materials Logging (Offline-First)

   Build the React Native + Expo screens for a technician's job-logging flow. Requirements:
   
   1. Job list screen: shows the tech's assigned open jobs for today, pulled from Supabase, cached locally for offline access.
   2. Job detail screen: a large [Start Timer] / [Stop Timer] button that writes a time_entries row locally first (using a local SQLite queue), then syncs to Supabase in the background when connectivity returns. Show a "queued, will sync" badge when offline.
   3. Materials capture: camera button opens the device camera, captures a receipt photo, uploads to Supabase Storage when online (queues locally when offline), and calls the Veryfi receipt OCR API server-side to extract line items and total cost. Show the extracted line items to the tech for a one-tap confirm before saving to the materials table.
   4. Drive time: request background location permission; auto-start a drive_segments entry when the tech leaves a geofence around the previous job site, auto-stop on arrival at the next job's address.
   
   Handle sync conflicts by last-write-wins on the local queue, with a visible "3 items syncing" indicator. Every screen must remain fully usable with zero connectivity.

3. 3. Owner Margin Dashboard + Underpriced-Category Flagging

   Build the owner-facing Next.js dashboard for MarginTrack.
   
   Main view: a live table of jobs (open and closed in the last 90 days) with columns: job_type, crew, quoted_price, actual_labor_cost (sum of time_entries * technician.loaded_hourly_rate), actual_materials_cost (sum of materials.total_cost_cents), actual_drive_cost (drive_segments.miles * $0.67 IRS mileage rate as a proxy), computed_margin_pct. Sort by margin ascending by default so the worst jobs surface first. Support filters by crew and by job_type.
   
   Rollup views: margin by crew (bar chart, this week vs last 4 weeks average), margin by job_type/week (trend line per job type over the last 12 weeks).
   
   Weekly flagging job (run via a scheduled Supabase Edge Function or Vercel Cron every Monday):
   - Query all job_types with 3+ closed jobs in the trailing 30 days
   - For any job_type averaging under a configurable margin floor (default 15%), call the Anthropic API with the job_type's cost breakdown and ask it to write a 2-3 sentence plain-English summary of why the category is underpriced and a suggested reprice range
   - Store the flag and summary, surface it as a dismissible banner at the top of the dashboard, and email/SMS the owner a digest via Resend/Twilio
   
   Use Claude with a strict JSON schema for the flagging summary: { job_type, avg_margin_pct, jobs_analyzed, likely_cause, suggested_price_adjustment_pct, summary_text }.