Qronos — protection for agents that act on their own
Shield protects the door. Qronos watches what happens behind it.
- — It does not handle network attacks or link flooding — that remains the WAF's job.
- — It's not Shield in different packaging: Shield guards an application's entry point, Qronos the behaviour of whoever acts inside it.
- — It's not monitoring. Monitoring records what already happened; Qronos stops the action before it executes.
Authorizing an agent once at the start isn't enough
Permission at the start of a task says nothing about where the agent will be several dozen steps later. The problem tends to be the whole, not any single step.
- 01The agent gets a taskFor example: prepare an overview and send it out. Permission is granted once, at the start.
- 02It breaks the task into stepsRead the data, sort it, prepare the messages, send them. Each step looks harmless on its own.
- 03The last step has a consequenceSending hundreds of emails, rewriting records in a system, or a payment. At this point, knowing the agent was authenticated at the start isn't enough.
- 04This is where Qronos decidesIt decides before such a step runs — by a rule you can point to, and with a record.
Qronos is in development. It isn't part of any plan, it can't be deployed today, and it doesn't appear in any price list. If you're interested, reach out and we'll let you know when it's ready.
Five things Qronos is built on
Open whatever interests you. We describe design principles, not a finished product.
Confidence at every step an agent takes
Qronos assesses each step on its own, so the agent stays within the limits you gave it.
Qronos assesses every step, tool and document on its own while keeping an eye on where the whole task is heading. That way it also catches the case where steps that look harmless one by one add up to something you never intended. The limits you set for the agent at the start hold until the very last step — so you can hand it work with peace of mind.
Qronos assesses every step, tool and document on its own while keeping an eye on where the whole task is heading. That way it also catches the case where steps that look harmless one by one add up to something you never intended. The limits you set for the agent at the start hold until the very last step — so you can hand it work with peace of mind.
A decision before execution
Qronos doesn't decide after the fact, but before the step.
An ordinary step can go through. A step that calls for closer attention is watched. A step with a real consequence, such as a bulk send, a write to a system or a payment, can wait for a person to confirm it. And a step that goes beyond what was allowed is stopped. The difference from after-the-fact review is timing: the step is decided on before it runs.
An ordinary step can go through. A step that calls for closer attention is watched. A step with a real consequence, such as a bulk send, a write to a system or a payment, can wait for a person to confirm it. And a step that goes beyond what was allowed is stopped. The difference from after-the-fact review is timing: the step is decided on before it runs.
Decisions you can evidence
Not just that something was blocked, but why.
Every Qronos decision points to a specific rule and leaves a record behind that you can present at a review. Whoever later asks why the agent wasn't allowed to continue gets an answer they can cite, not just an outcome.
Every Qronos decision points to a specific rule and leaves a record behind that you can present at a review. Whoever later asks why the agent wasn't allowed to continue gets an answer they can cite, not just an outcome.
Relationship to Shield
Shield protects inputs, outputs and actions. Qronos continuously watches the agent's run.
Shield sits at the application: it assesses what goes into it, what comes out of it and which actions are triggered through it. Qronos looks from the other side, at an agent that is already working, and throughout the whole run watches whether it goes beyond what you permitted, even when nobody would otherwise notice a deviation. You can deploy Shield today; Qronos, not yet.
Shield sits at the application: it assesses what goes into it, what comes out of it and which actions are triggered through it. Qronos looks from the other side, at an agent that is already working, and throughout the whole run watches whether it goes beyond what you permitted, even when nobody would otherwise notice a deviation. You can deploy Shield today; Qronos, not yet.
Architecture
Decision-making is a separate layer above the agent's steps, kept apart from the agent itself.
We're designing Qronos as a layered decision architecture. Keeping it separate from the agent is deliberate: whether a step may run isn't decided by whoever carries it out. Every decision is written to the audit log. The architecture is designed with the use of quantum computers in mind.
We're designing Qronos as a layered decision architecture. Keeping it separate from the agent is deliberate: whether a step may run isn't decided by whoever carries it out. Every decision is written to the audit log. The architecture is designed with the use of quantum computers in mind.
There are people behind the product you can reach by name
A company doesn't buy a security layer from an anonymous vendor.
