TorCyte
Sign up
Cognitive AI — patent-pending ISO architecture

AI that learns continuously. No retraining ever.

Torcyte is a cognitive API built on Indiscrete Scalar Ontology — a physics-grounded architecture that learns from every interaction in real time and provides uncertainty bounds with the rigour of conservation laws. It's the same engine that powers Aiva and Tor, our AI companions — available to build with directly.

Start free — 100 calls/month See pricing

Card required · First 500 calls free · £0.50/1k after · Cancel anytime

Looking for an AI companion instead of an API? Meet Aiva & Tor

TorCyte API ConsoleLIVE
// POST https://torcyte.ai/api/engine/search
X-API-Key:tc_live_••••••••••••3f9a
{
"query":"continuous learning without retraining",
"top_k":5
}
// 200 OK — 47ms
{
•iso_learning0.9812
•continual_adaptation0.9541
•online_learning0.9203
•cognitive_growth0.9014
•theta_accumulation0.8876
"omega_conserved": true, "confidence": 0.941
}
Patent-pending ISO architecture No PII stored — embeddings only Cross-modal semantic search Ω-conserved confidence bounds

A fundamentally different kind of AI

Not a transformer. Not an LLM wrapper. Built on ISO — Indiscrete Scalar Ontology — a patent-pending cognitive architecture rooted in geometric field theory.

Continuous learning

Every API call teaches the engine. Concepts crystallise from interaction, not batch training. Scale (θ) is monotonically non-decreasing — the system never forgets.

Meaning before words

ISO geometry is pre-linguistic. Concepts crystallise as geometric structure, not tokens — so the engine reasons over the relationships between ideas rather than matching surface text.

Physics-grade confidence

Combination Potential (Ω) is conserved like energy. Every response ships with uncertainty bounds derived from structural coupling density — not softmax probability.

Cross-modal semantic search

Audio, visual, and text embeddings share the same geometric space. A spoken query can retrieve visual results. No modality-specific indices.

Personalisation that persists

Sensory preferences accumulate across sessions. The engine learns what each user responds to in embedding space — no words needed.

No gradient descent anywhere

Zero backpropagation. Zero weight updates. Pre-allocated tensors only. Deterministic, auditable, and safe in regulated environments.

Six core endpoints

One API. Everything cognitive.

From semantic search to novelty detection to live personalisation — the full ISO cognitive stack behind a clean REST API. Average latency under 80ms on a single CPU worker.

POST
/v1/searchPhysics-grounded semantic retrieval
POST
/v1/noveltyPhysics-grounded anomaly score
POST
/v1/learnLive concept ingestion — no retraining
POST
/v1/similarCross-modal nearest-concept lookup
POST
/v1/personalisePer-user sensory preference bias
GET
/v1/confidenceΩ-conserved uncertainty bounds
TorCyte.aivantor.com / dashboard

Welcome Charles!

C

Total Points

33

Rookie · L1

Calls

9

3 this week

Avg Score

22%

needs work

Skills Insights

Confidence
41%
Objections
50%
Info Gathering
21%
Pitching
27%
Introduction
27%
Questioning
14%
Defining Need
20%
Closing
14%

From zero to production in minutes

01

Get an API key

Sign up, choose a plan, and your key is ready in 30 seconds. No sales call.

02

Submit concepts

POST text, embeddings, or structured data to /v1/learn. The engine crystallises concepts immediately.

03

Query and personalise

Use /v1/search, /v1/novelty, and /v1/personalise to build products that get smarter with every interaction.

04

Trust the confidence

Every response includes Ω-derived uncertainty bounds. Know how confident the system is before you act.

Ready to build with cognitive AI?

500 free API calls every month. Card required to activate — you’re only charged when you exceed 500 calls. Cancel anytime.

Start building free Read the docs