The AI Trust Company · Agentic trust, in glass

Trust for AI that acts.

AI agents are moving from answering questions to taking actions — sending, committing, deciding. Claire makes those agents provably trustworthy: governed at every gate, sealed in a tamper-evident record, and open to inspection. We hold ourselves to it first.

How GATE works See the record — we operate in glass
0
agent actions that ship without a human's approval
100%
of decisions sealed in a tamper-evident, hash-chained record
2
live installs you can inspect — not a demo environment
Art. 12 + 14
EU AI Act logging & human-oversight, mapped by design
The category

The last wave of AI answered questions.
This one takes actions.

An agent that drafts a reply is a convenience. An agent that sends it — that moves money, changes a record, or commits your company — is a liability, unless someone can prove what it did, why, and that a qualified human stood behind it. That proof is Agentic Trust. It is the bottleneck on every serious deployment, and it is the whole of what Claire builds.

Read what Agentic Trust means →

Product one

GATE — the Governed Agent Trust Engine

GATE is the operating system of trust beneath a deployment. Agents produce; a gate holds the action; a qualified human — never the one who submitted it — approves; the decision is sealed in a tamper-evident ledger. Autonomy is earned, gate by gate, never assumed. It works across agent frameworks and model providers.

01

Produce

An agent drafts an action — a message, a term, a record change. Nothing leaves the gate.

02

Gate

The action is held, scoped, and checked against the limits you set.

03

Approve

A qualified human — not the submitter — signs off, edits, or hands it back.

04

Seal

The decision is written to a hash-chained record: who, what, when, and why.

05

Earn

Autonomy is proposed only when the evidence clears the bar — and can be revoked.

How GATE works →

Who it's for

Built for regulated finance,
where a wrong move is audited.

GATE earns its keep wherever an agent's action carries real consequence and someone will later be asked to prove it was sound. Nowhere is that sharper than regulated financial services — lending, fintech, banking, insurance. It's the domain of our reference install, and it's where we start. You'll get the most from GATE if:

Your agents act, not just draft

Consequential actions

Approvals, borrower and customer communications, pricing and terms, servicing, collections — an agent that does something, not one that writes internal copy a human was going to rewrite anyway.

A wrong move is expensive

Money and compliance on the line

Funds move, a commitment is made, a limit is waived — and a regulator, auditor, or partner can later ask "who approved this, and why?" You need the answer already on the record.

You answer to someone

Examiners, auditors, a board

You operate where accountability is existential and evidence beats assurances. GATE's record maps to the logging and human-oversight standards your examiners already use.

You're past the pilot

Moving to production

You're scaling an AI workforce, and the trust layer is what stands between a promising demo and something you can actually let touch a customer or a ledger.

Where GATE is a smaller fit — said plainly

If your agents only draft low-stakes internal text, and nothing they do moves money, commits the company, or has to be defended later, GATE still works — but it's built to earn its keep where the stakes are higher. We'd rather tell you than oversell.

The proof an underwriter or examiner accepts in finance is what every high-consequence domain will need next — healthcare, legal, insurance operations. Finance is simply where the need is sharpest today. So it's where we begin.

Proof over promise

A trust company should be the
first thing it inspects.

We don't ask you to take governance on faith. Claire runs its own AI workforce under GATE, in glass — and so does LoanCirrus, a regulated fintech we also own, held to the same governance as any tenant. Two live installs you can inspect, not a demo.

Witness one — Claire, on itself

We run the governance we sell.

Claire's own sales and marketing agents operate entirely under GATE. Every draft, approval, correction, and revert is recorded. If our own governance failed, you would be able to see it — which is exactly the point.

Witness two — LoanCirrus, flagship

A regulated lender, under the same standard.

LoanCirrus, a lending-technology company operating across 25+ countries under common ownership, runs its AI agent workforce under GATE — held to the same standard as any other tenant, in a domain where audit and accountability are existential.

Disclosure

LoanCirrus and Claire share common ownership through their founder. LoanCirrus is a governed GATE tenant — not part of Claire — and receives no preferential treatment. We name the relationship in plain sight because a trust company that hides one isn't one.

See how we operate in glass →

Recognition, not invention

Built to the standards your
auditors already trust.

GATE's record maps to the logging and human-oversight requirements that regulators and procurement teams are already converging on — applied to high-risk-grade rigor by choice, ahead of requirement. We borrow recognition rather than inventing our own.

EU AI Act

Logging & oversight

GATE's hash-chained record maps to Article 12 (automatic logging) and Article 14 (human oversight) for high-risk AI.

ISO/IEC 42001

AI management system

The world's first AI management-system standard — increasingly a procurement precondition. GATE supplies the evidence layer.

NIST AI RMF

Govern · map · measure · manage

GATE operationalizes the framework's accountability lineage: every action traceable to a responsible human.

See the mapping — and what we do not claim →

Why "Claire"
"The name carries ai at its centre — embedded inside something human, discovered on second glance. And clair means clear. A trust company has exactly one product: clarity about what a machine did, and who stood behind it."
Michael Claire — Founder & Principal · Software architect · Operator of the reference install
Questions

Agentic trust is a new category.
Here's what people ask first.

What does "the AI Trust Company" mean?

We make AI agents provably trustworthy enough to act. Our first product, GATE℠, governs every agent action, records it in a tamper-evident ledger, and lets autonomy be earned rather than assumed. Trust is the product; software is how we deliver it.

What is GATE℠, exactly?

The Governed Agent Trust Engine. Agents draft an action; a gate holds it; a qualified human — never the one who submitted it — approves; and the decision is sealed in a hash-chained record of who did what, when, and why. It runs across agent frameworks and model providers. See how GATE works →

Who gets the most value from GATE?

Regulated financial services, first — lenders, fintechs, banks, and insurers deploying AI agents that take consequential actions: approvals, borrower and customer communications, pricing and terms, servicing, collections. GATE earns its keep where a wrong action moves money or triggers compliance exposure and someone will later be asked to prove it was sound. If your agents only draft low-stakes internal text, it's a smaller fit — and we'll say so. See who it's for →

Does GATE depend on a specific AI model — or on Claude?

No. GATE's governance is model- and framework-agnostic. It governs agents built anywhere, on any model, and the trust layer survives even if every model underneath is swapped. That independence is deliberate: a trust layer tied to one vendor isn't a trust layer.

What does "in glass" mean?

We operate the governance we sell, transparently. Claire's own agents and LoanCirrus — a regulated fintech under common ownership — both run under GATE, and we show how. A trust company should be the first thing it inspects. See the proof →

Is Claire compliant with the EU AI Act?

We map GATE to the Act's logging (Article 12) and human-oversight (Article 14) requirements and apply high-risk-grade controls by choice, ahead of requirement. We don't make a blanket compliance claim — obligations depend on your specific use case, and we'll tell you plainly where a control helps and where you need your own counsel. See the mapping →

Deploying agents that act?
Start with the trust layer.

Whether you're a lender, fintech, bank, or insurer installing an AI workforce or governing one you already run, it begins with a conversation about what your agents do and what proof you need. Founding engagements are open.

Book a conversation Take the 2-minute Scorecard
trustai@clairely.ai · Instagram · Toronto, Canada