Persona Protocol
The 7-layer architecture for contextual AI identity.
A specification for identity as a record rather than a prompt. Seven layers, from a controlled vocabulary of psychological attributes up to the sentence a reader actually hears — each one addressable, versioned, and able to change under pressure without ceasing to be the same entity.
- PP-SPEC-001
- 2026.8
- Working draft
- 2025-12-04
Composed, not enumerated per character
Substrate through expression
Voice, craft, worldview, psyche, shadow, relation
USPTO, priority established
Scope
What the protocol coversGenerating a paragraph in a voice is solved. Generating the ten-thousandth paragraph in the voice of someone established in the first is not, and the reason is structural: the identity lives in the context window, so it degrades exactly as fast as the window fills.
Persona Protocol moves the identity out. An entity's traits, history, relationships, and speech habits live in a structured record that each generation pass reads from and writes back to. The model is re-grounded against that record instead of being asked to re-read the draft. What the protocol specifies is the shape of that record, the rules for inheriting between its layers, and the operations that are allowed to change it.
In scope
- The layer model and its inheritance rules.
- Attribute composition, and what a valid composition is.
- Mutation under narrative force — direction and magnitude, not rewrite.
- Snapshot and replay, so state at any point is reconstructible.
- Drift measurement against the canonical record.
Out of scope
- Model selection, routing, and inference. The protocol is provider-agnostic by construction.
- Prompt engineering. A prompt is an output of the record, never the record.
- Claims about consciousness or sentience. The protocol describes a data architecture.
Position
Relationship to Narrative GeometryPersona Protocol is the identity engine underneath Narrative Geometry, a long-form manuscript platform built on it. The platform is the protocol's first and most demanding consumer, not its purpose: if an identity holds across 150,000 words of fiction — where every inconsistency is visible to a reader and nothing is forgiven — it will hold across shorter and more forgiving work.
The two are developed separately and on purpose. Narrative Geometry is a product with authors, pricing, and a beta queue. Persona Protocol is the specification and the filings, and it applies anywhere a synthetic entity has to stay itself across time — support agents that remember their own commitments, tutors whose manner is stable across a term, characters in interactive work, any long-running system where drifting into a different personality is a defect rather than a feature.
Both are operated by Alalonde Corporation in Ottawa.
The seven-layer semantic model
L7 substrate → L1 surfaceRead the table downward. L7 is the substrate — the vocabulary everything else is written in. Each layer below inherits from the one above it and may narrow or override it; none may contradict L7. L1 is the only layer a reader ever encounters directly, and it is derived rather than authored.
| Stack | Layer | Function | Persistence | Mutability |
|---|---|---|---|---|
| L7 | Ontology | The controlled vocabulary. Every trait, tone, wound, fear, and attachment style the system can express, with the behavioural instruction attached to each. | Permanent | Versioned |
| L6 | Narrative personality | 51 attributes composed into a coherent psychological profile. Portable across projects; owns no name and no history. | Permanent | Author-edited |
| L5 | Persona | A narrative personality bound to goals, guidelines, and a purpose. The reusable working identity. | Permanent | Author-edited |
| L4 | Character | A persona instantiated with a name, a body, and a history inside one work. The first layer that can be pointed at. | Per project | Under narrative force |
| L3 | Context binding | Situational adaptation. The same character in a funeral and a negotiation resolves differently without either resolution being a different character. | Per scene | Continuous |
| L2 | Session state | Working memory for the current pass: what has just been said, what is unresolved, what the character currently believes. | Per session | Continuous |
| L1 | Expression | The surface. Diction, rhythm, tell, silence — the only layer a reader ever encounters directly. | Transient | Derived |
Persistence is how long an instance of the layer outlives the session that made it. Mutability is what is permitted to change it.
Patent ledger
United States Patent and Trademark OfficeFour provisional applications covering the computational identity architecture. A provisional filing establishes a priority date. These are not granted patents, and nothing on this page asserts that a patent has issued.
Computational Identity Architecture for Persistent, Adaptive, and Context-Aware Artificial Entities
The base architecture. A layered identity record that survives independently of any session, model, or provider, and can be re-grounded against rather than re-read.
Computational Identity Lifecycle Engine — System and Method for Computational Identity Mutation, Snapshot Serialization, and Temporal Persona Evolution
The lifecycle engine. How an identity is snapshotted, advanced, and replayed, so that the state at any point in a work is reconstructible rather than inferred.
Computational Identity Lifecycle Engine — Force-Aligned Identity Mutation, Snapshot Serialization, and Temporal Persona Evolution
Mutation under pressure. Narrative force is applied as a vector against the attribute space, so change has a direction and a magnitude instead of being a rewrite.
Drift-Coherence Measurement and Correction for Long-Horizon Synthetic Identity
The measurement half. Quantifies how far an identity has moved from its canonical record and distinguishes earned development from decay, which is the difference between a character arc and a continuity error.
Status
What is built, and what is not- Seven-layer record, in production under Narrative Geometry
- 51-attribute composition across six cognitive domains
- Snapshot serialisation and temporal replay
- Mutation under narrative force
- Public specification document
- Third-party API and reference implementation
- Drift-coherence metric, published and reproducible
- Conformance suite
Items on the right are not scheduled. Ask about any of them and we will tell you whether there is a date.
Enquiries
Licensing, research, implementationIf you are building something where a synthetic entity has to stay itself across time, describe where it currently falls apart. That part tells us the most.
contact@personaprotocol.ai · Alalonde Corporation, Ottawa, Ontario