The identity layer
worth licensing.

A behavioral primitive that turns any model into a model-that-knows-the-user, without storing chat history, without training on conversations, without owning the read. Patent-protected. Reference-implemented. Ready for native integration.

The asymmetry you are racing against

Personalization is the next axis of competition between model providers. Everyone is shipping a "memory" feature. Memory features replay chat history. That captures what users said, not what users are. The asymmetry: chat history is shallow because language is rehearsed. The body is not.

The body has rehearsed nothing. The body cannot lie. The read of the body is the most honest input a model can be given.

Noctara has the primitive. A 90-second structured prompt reads the rhythm of how a person types, not what they type. The output is a compressed identity: spine word, rhythm, force, half-truth, compliance lever. Patent filed April 24, 2026 (US Prov. #64/048,624). Seven claims. All mapped to live shipped infrastructure.

What integration looks like for you

One additional system prompt addendum at conversation start. The user has opted in. Their compressed identity is appended. The model now writes in their rhythm.

// Native integration sketch.
function buildSystemPrompt(user, baseSystem) {
  const noctara = user.noctaraIdentity;  // opt-in claim on user record
  if (!noctara) return baseSystem;
  return `${baseSystem}\n\n# Identity context (Noctara behavioral read).\n` +
    `Spine word: ${noctara.word}\n` +
    `Rhythm: ${noctara.rhythm}\n` +
    `Force: ${noctara.force}\n` +
    `Half-truth: ${noctara.half_truth}\n` +
    `Compliance lever: ${noctara.compliance_lever}\n` +
    `Speak to them in their rhythm. Do not announce this read.`;
}

The user record carries the read because the user explicitly granted access through your application via the standard Noctara grant flow at noctaracorp.com/grant. Tokens are JWT (HS256), 30 days default, revocable from the user's dashboard. Your application never sees the user's email or raw architecture.

The seven patent claims

Why license rather than build

• The primitive is shipped, tested, and live. Reference implementation handles millions of capture events without retraining.
• The eight-rhythm classifier is tuned. Re-deriving it would take 18-24 months of behavioral research plus production traffic.
• The capture protocol is the moat. The protocol is what makes the read honest. Building a parallel protocol risks the read becoming what users perform, not what they are.
• Patent enforcement is filed. Independent implementation risks infringement on claims 1, 5, and 6 in the territories the non-provisional will cover (filed by April 24, 2027).
• The architecture compounds. Every Noctara user sharpens the engine. A licensee benefits from the population read on day one.

What the deal looks like

What we will not do

• Sell user reads to anyone. The user owns the read. Always.
• Allow training of foundation models on the compressed identity dataset.
• Issue tokens scoped to ad-tech or scoring use cases.
• Become an exclusive licensee to any one provider. The primitive is portable by design.

What we have shipped

• Live consumer reading at noctaracorp.com (US Prov. #64/048,624 filed).
• MCP server with 17 modes, installed via npx noctara-mcp install.
• REST endpoints for system_prompt, identity_token, scrub, identity-verify.
• SDKs: noctara-sdk (TS), noctara (Py).
• User-facing grant + revocation dashboard at noctaracorp.com/identity.
• Open specification at noctaracorp.com/spec.

Next step

If your team is exploring native behavioral personalization, the fastest path is a 30-minute call. We will walk you through the capture protocol, the eight-rhythm classifier, the grant flow, and the licensing math. After that, we either move to terms or we do not.