Pay for proof, not another dashboard.

Start with what you have. Import your existing model inventory spreadsheet as a CSV in an afternoon. You review a preview first, so nothing goes live until you've approved it. Committee onboarding is one calendar invite, added once. There is no forced data migration and no new system your team has to learn from scratch. The workflows deliberately mirror the email, calendar and spreadsheet habits you already have. For design partners, we're hands-on through all of this, with a direct line to the founder rather than a support ticket queue.

More than a mature vendor's usually are. We revisit design partner terms together as the product matures, so you won't be locked into a multi-year rate that assumed features we hadn't built yet. We'll sign a standard MSA and a data processing addendum for your counsel to review. We're also open to shorter initial terms, defined exit points, and wording that scales with what has actually shipped versus what is still roadmap. Our public capability status document exists so that conversation can rest on what the product does today.

No. A proof sprint runs for 30 days on one production workflow you pick, at a fixed fee agreed before it starts, and that fee is credited in full against the first year if you go ahead. A bank or larger institution can scope a 90-day pilot on the same terms instead. A paid pilot gets your team's time and a real decision at the end. If you only want to look around first, the Free plan puts one model or agent on the real record at no cost.

Yes, and it is scoped on a call rather than priced from this page. It is a fixed-fee engagement around the one order you are working under, and it needs an Enterprise licence behind it. Coriqo records what happened and who signed. A named person at your bank decides whether a control is fixed, and your regulator decides when an order closes. Talk to us.

A regulated buyer should ask an early-stage vendor this, and the question shaped how we built the product. Verification doesn't depend on our servers. Merkle proofs and Ed25519-signed checkpoints can be checked offline with a public key alone, so your evidence record isn't stranded if Coriqo the company disappears. On the commercial side, we're open to writing data portability and export guarantees into the MSA for anyone who needs them in writing before signing. That includes a full export in open formats (CSV, JSON, the raw hash-chained event log), on request or at contract termination. Ask for it. It's a reasonable ask and one we expect from any vendor in this space.

Free is community support (docs and email, no response-time commitment). Pro gets direct email support, typically same or next business day. Firm adds priority handling for your whole team. Enterprise adds a named technical contact and a defined SLA with response-time commitments, set during the scoping conversation to match your deployment (a managed cloud instance and an air-gapped on-premise deployment need different support).

Yes. The homepage has a live interactive demo with illustrative data you can click through right now (inventory, governance lifecycle, lineage, and generating an examiner package), no account needed. For a deeper look, including a real security review packet, email us directly and we'll set up a walkthrough.

Two options today. Managed private cloud: each workspace gets its own separate section of the database (a Postgres schema, t_yourbank) rather than sharing tables with other customers. Every request is limited to that schema centrally, so your separation doesn't depend on every query remembering to add a filter. Customer-hosted, in your own AWS or Azure account, for +50% on the plan: Coriqo needs no outbound network connection to run day to day. One caveat, stated plainly: we can optionally post a cryptographic checkpoint to a public log (Sigstore's Rekor) so someone outside our control can vouch for it. That step needs outbound access. In a restricted network it waits as "pending" rather than failing, and you can retry it whenever a connection is available. Nothing about core operation depends on it.

No. Zero-retention by design. Your data is never used to train any model. Each customer has a separate Postgres schema, so there's no shared table your records could leak across. Optional AI features (drafting, narration) send each request to a model provider, and neither we nor they train on your data under our agreement.

On top. Coriqo is read-only by design. It watches and records. It never sits in the path that serves your models, so your models never depend on it to run. Your committee still meets by calendar invite and email, and adding Coriqo's address to that invite is the whole onboarding step. Your existing inventory comes in by CSV or Excel import, with a real preview before anything is saved and a second person approving it. There is no forced migration.

Yes. Vendor models are most of what a financial institution governs. Attach the vendor's validation report and model card to the exact version it covers, locked to that version so "which document did we rely on" has one answer. Log your own user-gap analysis as tracked findings with owners and due dates. Set a vendor due-diligence review date once. Coriqo reminds the owner, flags it when overdue, and shows it in board reporting automatically.

Being direct, because it's easy to oversell: CSV and Excel import works today, with automatic format detection, a real preview before anything is saved, and rollback with dependency checks. Evidence Connect pulls evidence from the systems marked Available on its page: Jira, GitHub and SharePoint today. S3, Snowflake and MLflow are not there yet. Loading the model inventory itself from a live system is still a stub in our code, so the inventory comes in by CSV or Excel. If a live connector is a requirement for your evaluation, tell us. We discuss it openly with design partners.

Every governance event (an approval, a finding, a committee decision) is linked to the one before it by a fingerprint, called a hash. Change any record afterwards and the link breaks, which shows up straight away. On top of that, events are grouped into a Merkle tree and signed with Ed25519. An examiner can download a small proof for any single event and check it with one command on their own machine, offline, using only a public key. They need no Coriqo account, no network connection and no trust in our servers. Checkpoints can also be posted to Sigstore's public Rekor log, the same kind of append-only log browsers use to police TLS certificates, so a witness outside our control can vouch for them. We publish the exact tooling (verify_proof.py) rather than asking anyone to take our word for it.

Honestly: the hash chain makes tampering detectable, not impossible. A database administrator with direct database access could still delete the most recent rows. That breaks the chain, but it only shows if someone runs the verification check. That's why we add Merkle checkpoints and optional third-party anchoring. They move the guarantee from "trust our database" toward "trust the math." Our public roadmap (Coriqo V2) adds independent outside witnesses who co-sign checkpoints, to remove single-party trust entirely.

It means the governance workflow (draft, independent review, validation, approval, production, retirement) mirrors what those frameworks expect an AI governance program to evidence. Specific checks are built into the workflow itself rather than left to people's memory: segregation of duties between owner and reviewer, no self-approval, and no approval over an open critical finding. We're precise about the boundary. Gap detection is real and runs against your live data. The regulatory framework labels attached to a gap are a curated cross-reference for readability, not a rules engine that parses regulatory text. For example, a gap raised under SR 11-7, superseded by SR 26-2 in April 2026, keeps its OCC 2011-12 cross-tag, because that guidance mirrored it. Coriqo does not provide legal or regulatory advice. It provides the evidence structure your own compliance program uses to demonstrate its governance practice.

Record them. That boundary is deliberate. AI features across the product (examiner reply drafts, committee minutes parsing, challenge-flag narration) use your own data and only produce drafts. Nothing enters the tamper-evident record until a named person approves it, and the record always shows that person as the actor, never "AI." Coriqo works out vote tallies and quorum from the attendee data you enter, as a check on your count. The decision itself (approve, reject, approve with conditions) stays a human judgment call, on purpose.