QUAICU
How every delivery
stays checked.
For technical reviewers. The detail behind our promise that nothing ships unchecked: how oversight is built into the way we work.
Rules handle what must not vary.
AI helps where judgment is useful.
Fixed rules.
For important actions where mistakes are not acceptable. The same input always gives the same result, and AI is not allowed to override them.
AI within limits.
For predictions, risk flags and draft content: always reviewable by a person, never run on its own.
Both run under the same Governance Kernel. People have the final say on anything important.
Four layers. One set of rules for all of them.
The platforms and workflows we build and support for you, across web and mobile. The same governance foundation sits underneath every engagement.
Where your non-negotiables are written down. They are checked on every action, give the same result every time, and are designed so no one can quietly override them.
AI drafts and suggests within limits, and a person can review everything it produces. AI proposes; your rules and your approvers decide what runs. It runs nothing on its own, except fixes you pre-approve under out-of-hours cover.
On-premise, private cloud, hybrid, or hosted by us on AWS in India. Each client's data stored separately. Sign-in and access management. Encryption at rest and in transit. An audit record designed so it can't be edited after the fact.
Every change in everything we deliver is designed to pass through the same checks.
Every action is written to one audit record. Each entry is linked to the one before it, so the record is designed so it can't be edited after the fact, and it can be replayed step by step. Critical rules are designed so no AI agent, integrator or privileged user can bypass them. Because everything is in one record, we can produce a full audit report when you ask.
With out-of-hours cover, fixes you pre-approve, within rules you agree with us in advance, can run on their own; a person reviews each one the next working day. Everything else waits for a person.
Every AI action is designed to follow one sequence.
Who handles your data
on our behalf.
These providers handle client code, requests or messages while we do your work. We list them so you can see where your information goes.
- Anthropic
- AI model provider for our coding agent, including its cloud sessions. Sees the code, request details and outputs for the work in hand. Used under business terms that do not allow training on your content. Standard retention is 30 days.
- OpenAI
- AI model provider for our second coding agent, including its cloud tasks. Sees the code, request details and outputs for the work in hand. Used under business terms; business data is not used for training by default.
- Meta (WhatsApp Business Platform)
- Carries requests and status messages. Sees phone numbers, names, message text and attachments.
- Amazon Web Services (India region)
- Hosts our internal system, each client's request bot, and the software we host for clients.
Where each provider processes data is set out in its own terms and confirmed in your agreement. We update this list when providers change. Last reviewed November 2026.
Guides from QUAICU Notes.
The technical depth continues on kernel.quaicu.org.
API reference, Python SDK guide, CEL policy language, deployment runbooks, and the architecture decisions behind every governance layer.