What Nage does.
Nage develops KLM, a candidate standard for recording the declared basis of AI outputs, and operates its reference implementation. SEDIM supplies the architecture; Nage AI OS carries records across services.
Build the basis
for a better review.
We develop infrastructure for source-aware AI outputs and records that another party can inspect within a stated scope.
A practical trust anchor
A shared record, without a blanket guarantee.
Our trust-anchor proposition is practical: give applications, reviewers and assurance tools a shared record to inspect.
The record should expose source identity, signal status and verification scope without claiming more than its evidence establishes.
A disciplined separation.
Current services, architectural intent and research are different commitments. The current SEDIM grounding service uses Qwen3-8B. Trained VARVE weight injection, Craton and Craton Pro remain development paths.
The research belongs in the open conversation.
A failed support baseline, incomplete provenance history or pre-acceptance model path should remain visible. Publishing those boundaries makes an evaluation more useful.
Meet us through a concrete question.
Bring a workflow, a source set and a review decision. We can discuss the record and evidence an evaluation would need to produce.
Continue the conversation
Start with something concrete.
Choose the visitor guide for a conference conversation or take the six-page company introduction with you.