RepairSnap: AI-powered front desk for auto body shops.
Body shops were losing jobs to voicemail. Downshift built the AI agent that answers every call, text, and chat in 30 seconds with a photo-based estimate.
Live at repairsnap.cc · canonical domain repairsnap.ai.
Screenshot from repairsnap.cc, captured 2026-05-06.
The problem
Auto body shop owners lose jobs because they can't answer the phone fast enough. Insurance estimators, customers, and tow yards all call during the same hours the shop is busy fixing cars. Every missed call is a potential job that walks to the next shop on the list.
The status quo answer is a paid receptionist or a callback service. Both add cost without solving the speed-to-quote problem. Customers want a price, not a callback. Insurance work wants a structured estimate, not a voicemail.
The ICP is the body shop owner who cares about not losing jobs, saving time, and making money. They don't care about AI. They care about the next $4,000 repair not walking out the door.
What Downshift built
A 24/7 AI front desk shops embed via one script tag. Twelve named features, each tied to one job a real body shop has to do today.
24/7 Front Desk
Handles calls, texts, and website chat continuously. Customers get an answer at any hour.
Photo Estimates in 30s
AI vision analyzes damage images and returns a cost range in 30 seconds. Backed by 39 vehicle databases and 150+ scraped market data sources.
Appointment Booking
Books the customer in automatically with confirmation. Configurable deposits ($25 to $200) collected at booking.
Lead Dashboard
Centralized view of every lead with photos, conversation history, and damage details.
Live Repair Updates
7-stage repair timeline with automated customer texts at each step. The shop never has to chase status updates.
Automated Follow-ups
Day 1, day 3, and day 7 sequential messages to unbooked leads so no one falls through the cracks.
Supplement Automation
Auto-generates insurance supplement requests when the repair scope changes mid-job.
Outbound Campaigns
Six ready-made marketing message templates the shop can fire to its existing list.
Conversation Scoring
Performance analysis with coaching suggestions for the human operators who do step in.
Screenshot from repairsnap.cc, captured 2026-05-06.
The architecture
Production-grade from day one. Three live services on the same monorepo, every layer typed end to end.
Core stack
- Bun monorepo (5 packages)
- Fastify 5 + TypeScript backend
- React 19 + Vite 7 + React Router 7 + Tailwind 4
- Astro 5 SSG for marketing
- Drizzle ORM + PostgreSQL
- BullMQ + Redis for jobs
AI surface
- OpenAI GPT-4o Vision for damage detection
- Anthropic Claude for repair quote synthesis
- 39 vehicle JSON databases
- 150+ scraped market data sources
- 51 state-level labor rate datasets
- Custom TypeScript eval loop with versioned prompts
Production services live
- repairsnap.ai (marketing site)
- app.repairsnap.ai (shop dashboard)
- api.repairsnap.ai (estimation API)
Ops
- Cloudflare R2 for media (shared bucket, prefixed)
- Stripe billing wired
- Google OAuth wired
- 151 marketing pages live
Real outcomes
Numbers below are from RepairSnap's live marketing on repairsnap.cc and from the production codebase. No fabricated metrics.
Time from photo upload to repair estimate.
Setup time to embed RepairSnap on a shop's existing site.
Estimated gross margin on the average $4,730 repair job recovered.
Of $149.50/mo service paid for by a single repair job recovered.
Vehicle databases driving the estimate engine.
State-level labor rate datasets.
Scraped market data sources behind the cost ranges.
Marketing pages live on the production site.
Sources: numbers above are quoted directly from repairsnap.cc (pricing, $2,130 margin example, 30s, 48h) and from the production codebase (39 vehicle DBs, 51 state datasets, 150+ scraped sources, 151 marketing pages). Specific shop-level revenue and lead volume are shared on request, not on a public marketing page.
Screenshot from repairsnap.cc, captured 2026-05-06.
What's in flight
RepairSnap continues to ship. The estimation engine moves through versioned prompt evaluations on a custom TypeScript eval loop, with strict accuracy criteria gating every release. The same engine is exposed as a developer API; FleetSnap is the first external consumer using it for fleet damage assessment.
Want this for your venture?
Downshift partners with non-technical founders to ship production-grade AI products in 3 to 6 weeks. Fixed scope, no equity.
See more shipped Downshift ventures on the wins page.