Showcase · live in production

Offlot.
From VIN to a finished campaign, with an AI assistant in the app.

A dealer enters a VIN, uploads photos and answers five questions. Offlot returns finished campaign copy for six channels, a calendar and the leads that come back. The AI assistant in the app creates a vehicle and its campaign after a clarifying question, and you confirm before anything is saved. Live at offlot.app, built end-to-end on Kumiko.

What this build proves

  • →AI is part of the product flow: VIN photo reading, number-plate detection, channel copy and the in-app assistant are wired into the dealer workflow.
  • →The assistant asks, you confirm: The AI assistant (Enterprise) asks what is missing, proposes the vehicle and the campaign, and a person approves before it is written.
  • →One vehicle, six channels: The channel catalog carries format, length, tone and image rules, so every post is shaped for the channel it is meant for.
  • →Every change is an event: Vehicles, campaigns and approvals are stored as events, so the history of a campaign is part of the data.
  • →Billing with Stripe and Paddle: Plans and credit packs are bought through Stripe or Paddle, with the consumer-protection steps built into checkout.
Offlot campaign composer: a campaign with AI-generated posts for several channels, approved per post
See the Offlot site →Read the docs

Enter the VIN, Offlot recognises the car, and the campaign copy for every channel is ready.

offlot · Today
Offlot Today screen with the vehicles and campaigns that need attention

The Today screen shows what needs a decision, from vehicles to campaigns.

offlot · Assistant
Offlot AI assistant creating a vehicle and its campaign after a clarifying question

The assistant sets up a vehicle and its campaign, and you confirm first.

offlot · Vehicle intake
Offlot vehicle wizard showing decoded VIN facts for review

A VIN becomes decoded vehicle facts that the dealer checks.

How it is built

Offlot owns the domain logic. Kumiko supplies the repeatable platform.

VIN photo + vehicle intake

AI output becomes structured vehicle data

The VIN-photo step returns a validated VIN, which feeds the vehicle intake flow. VIN facts are combined with dealer details, photos and publish checks.

AI document workflows →

Plate detection + image review

AI output stays reviewable

Number-plate detection returns regions for review. A manual review step sits between detection and any public use of the image.

AI pipeline →

Channel copy + prompt store

Different content for different channels

Channel copy is generated from structured vehicle facts and channel rules. Prompts come from a tenant-scoped, versioned prompt store and can be tried in the AI dry-run screen.

Prompt store →

In-app AI assistant

Create a vehicle and its campaign by conversation

The assistant (Enterprise) asks a clarifying question, proposes the action and a person confirms it before anything is saved.

AI assistant →

Provider adapters

The AI provider is replaceable

The app mounts an Anthropic adapter and an OpenAI-compatible adapter. Provider and model policy live in the AI foundation and pipeline layers, not in domain handlers.

Provider adapters →

Kumiko bundled-features

The platform work is composed in

Auth, sessions, user and tenant identity, file storage, jobs, secrets, notifications, data rights, retention, tiers, caps and billing come from the platform, next to Offlot's own features.

Plans and checkout →

Where the effort goes

The app owns the domain: vehicles, campaigns, channel rules, plate review, public vehicle pages, leads and the dealer screens. Identity, tenancy, files, jobs, notifications, tiers and billing come from the platform. The Offlot app holds more than 20 app-owned feature modules and more than 400 test files, including more than 140 integration tests (measured and rounded down).

Continue in the docs

Build your own app with Kumiko.

The docs take you from the first command to a running app. The Show Pony tutorial walks through the whole path on a real application.