Kantia KORA
Kantia’s agent platform.
KORA is our platform for putting AI agents to work within a framework: what they are allowed to do, what a person must approve, what is recorded. It is designed to mask personal information before anything is sent to a model. We are building it first for our own products.
StatusIn development
An AI agent with no framework is a risk. With one, it becomes a colleague you can supervise.
AI agents can now chain actions: read, draft, classify, prepare a decision. In a regulated profession, the question is not what they can do, but what they are allowed to do, and how you show it afterwards.
KORA is the answer we are building: a platform where each agent has a defined role, where the actions that matter go through a person, and where what was done can be consulted.
Who it is for
- You want to hand tasks to AI agents without losing control of what they do.
- You must be able to show who did what, and on what basis.
- You do not want to depend on a single model provider.
Designed to
Masked personal information
Replace personal information with substitutes before anything is sent to a model, then restore it on return; if masking cannot be done, nothing is sent.
Agents with a limited role
Give each agent a defined role and defined tools, and nothing else.
Human approval
Hold the actions that matter until a person approves them.
A history you can consult
Keep what the agent did, based on what, and who approved it.
Intended benefits
Delegate without losing control
The agent prepares; the decision stays with a person.
Be able to account for it
A history of what was done, based on what, and approved by whom.
Not tied to one model
KORA is designed to switch models without changing how you work.
The principle
You set the framework
The roles, the permitted tools, what requires approval.
The agent prepares
It carries out the permitted tasks and proposes the rest.
A person decides
They approve, correct or refuse; the decision is recorded.
Our design choices
The decision stays human
The actions you designate wait for a person’s approval.
Calculation is not left to the model
What must be exact is calculated by tools made for it, not estimated.
Tried on our own products first
We are building it first for our own products, before offering it more widely.
Frequently asked questions
- Where does data entrusted to KORA go?
- Masking is done in Canada. The masked text is then processed by the model provider, most often outside Canada: processing by the model therefore does not take place in Canada.
- Is masking perfect?
- No. Masking is measured, and we do not promise complete detection.
- Does KORA depend on one model provider?
- No. KORA is designed to run its own pool of models and to switch between them without changing how you work.
Your case deserves thirty minutes.
A conversation to understand your context and tell you, frankly, whether we are the right choice. The diagnostic that follows is free of charge.
Book a call