How it works
Small models, made to disagree
Under the hood is NSED — Peeramid Labs’ async deliberation runtime. Instead of one large model answering alone, specialised agents (Claude, Mistral, Llama, your own local models) reason in parallel, conflict, debate, and converge. Disagreement is the signal: where they diverge is exactly where a single model would have quietly guessed.
You get a single answer with the receipts — who reasoned, how well, on what evidence — not a black box you have to take on faith.
Compliance
The audit trail is a by-product
Regulated teams fail with black-box AI because they can’t show their work. Noolog’s oversight, dissent, and kill-switch traces are generated by the protocol itself, then mapped to the frameworks you answer to:
See the full regulatory mapping →
Deployment
Hosted, or on your own metal
Call the hosted API at api.peeramid.xyz and ship in an afternoon — zero infra. Or run it on-prem on your own hardware, where keys and context never leave the box: from a single Rack for a clinic or a firm, up to datacenter-tier edge. Same protocol, same audit trail, your choice of where the metal lives.