Three products.
Three places to start.
Each product is documented on its own, because each can be bought and used on its own. These pages describe what you can integrate with, observe and rely on — including where each product stops.
Compile-Time Inference
What it does, how knowledge is prepared and kept current, how missing evidence is handled, and how to evaluate it.
How you start: Assisted evaluation — scope, sources and success criteria are agreed with you first.
Read the Compile-Time Inference docsGate
Connect Gate, read a verdict, act on findings, and add it to pull requests.
How you start: Self-service — create an account, add the connection to your AI tool, and run a check.
Read the Gate docsHelix
Where Helix runs, the API your applications call, and how to use confidence signals.
How you start: Licensed deployment — a container image you run in your own environment.
Read the Helix docsUseful to evaluate. Not a recipe to copy.
Public documentation explains supported tasks, interfaces, operating behaviour, limits and how to run a fair evaluation. It does not publish the internal methods that make the products work. Approved evaluators can request deeper detail under an appropriate agreement.
Looking for proof?
See which measurements support which claims, and what each one does not establish.
Evidence & evaluationComparing with RAG?
Plain-language guides to where retrieval helps, where it falls short and what to test.
RAG vs Compile-Time InferenceSecurity questions?
Deployment boundaries, data handling and how to reach the security team.
Security information