The loop
A live database the AI actually operates.
UnitCheck reads your live operating data — every CODECO, lease line, per-diem accrual and yard move — reasons across the whole operation, prepares evidence-backed work like a D&D dispute pack or a repositioning batch, and waits for a person to approve before anything hits the record.
Live state
The AI reads your current tenant-scoped data — fleet, gate moves, leases, EDI, billing, yard, funds.
Across domains
It reasons over the whole operation at once: what is idle, mis-billed, overdue, or out of position.
Evidence-backed
It drafts the next action with a rationale and the records that justify it — not a guess, a citation.
Human in control
Nothing hits the ledger until a person approves. Every write is audited and reversible.
Walk your yard. It's a live digital twin.
This is YardVision — the walkable digital twin of a depot yard, drawn live from the UnitCheck database. Search a unit and fly to it. Walk the stacks. Arm planning mode and drag a restow — the same engine, physics, and audited planning loop operators use in the product.
YARDVISION // SCHEMATIC
Your fleet on every lane,
one audited loop.
The same core that draws the yard above carries your fleet's global flow — gate moves, EDI, per-diem, repositioning. Each chip on the globe is a platform event playing the loop: proposed by AI → approved by a person → executed & audited.
Your database. Your model.
Bring your own Anthropic or OpenAI key. The model that writes runs on whatever you pick, and when a better one ships you switch with one setting. Use a frontier model for hard reasoning, a leaner one to keep costs down. You hold the key, the spend, and the governance — while the model that reads, our own domain-tuned retrieval, comes built in.
Container trade runs on software built for a different era.
Most operators still run on transaction-era software: single-tenant, no real API, and no AI past a help-desk bot. These systems are capable and deeply entrenched. None of them can be run by an agent without a rebuild. UnitCheck started there.
- Owned, fine-tuned AI — not a help-desk bot bolted on
- Whole domain exposed as typed API + MCP tools
- A goal-directed agent that prepares audited batches of work
- Event-sourced — every action becomes an event
- Multi-tenant with database-level RLS isolation
- Every write gated by a human-audited resolution layer
- One core spans international maritime + US domestic
- Canonical double-entry posting, not sync-and-export
The real product, running here.
Not a mockup — these are real screens from the UnitCheck app, running on a full demo fleet. One operator console, recomposed for Operations, Leasing, Sales, Finance and Investor over a single source of truth. The Repositioning lens — the engine that turns idle steel into recovered cash — gets the longest look.
Real screens from the UnitCheck operator app, running live on a representative demo fleet.
Nobody else spans both.
Flip the segment flag and the same screens recompose for the other side of the business — the way the product works inside. Maritime and US storage run on the same engine and the same ledger, reached through the same API.
Enterprise controls are part of the product, not a footer.
UnitCheck has to earn trust from operators, controllers, fund admins, and investors. The controls below sit in the operating workflow itself: isolation, approval, ledger integrity, event history, and evidence.
Questions, answered.
What does early access include?
A guided pilot on your own workflows — we stand up a tenant with your data and walk it through with you. No commitment, and nothing to install.
What does it cost?
Pricing is being finalized with our early-access partners. Requesting access costs nothing and commits you to nothing.
How does our data get in?
UnitCheck reads the EDI you already exchange — CODECO, COPRAR, COPARN — and brings your fleet and contract records in through a typed API. Getting your data in is part of the pilot.
Is our data secure?
Every tenant is isolated at the database level, every AI-prepared change is approved by a person and audited, and AI inference runs on your own model account under your governance.
When do you launch?
Soon. Leave your details and we will reach out before launch to set up your pilot.