How does Agent Persona recognize the same person in a fresh chat?

The system compares a new conversation against existing persona graphs and explains the match using evidence like recurring preferences, communication texture, and task behavior.

What can teams use Agent Persona for?

Teams can use Agent Persona for continuity in AI products, persona-grounded testing, interview analysis, adaptive experiences, and clearer personalization.

Memory is not identity

Recognize the same person in a fresh chat.

Q: What is Agent Persona?

Agent Persona is a product direction and prototype system for turning conversations into stable persona graphs across memory, character, and skill.

Agent Persona converts conversations into persona graphs across memory, character, and skill so AI products can preserve continuity, explain the match, and reuse that identity for testing and personalization.

Request pilot access Read blog posts Star the repo

Core demo Talk to five people, open a fresh anonymous chat, then match it back to the right persona.

Identity layers Memory, character, and skill. Not just remembered facts and profile summaries.

Why now AI products are moving to longer-lived relationships, but continuity still breaks between sessions.

Recognition sequence

Fresh anonymous chat

“I always notice when a product forgets my priorities between sessions.”

continuity-sensitive
high recall
strategy-first

Persona graph engine

041

memory

character

skill

same-person confidence 94%

Winning evidence

stable preference stack across sessions
repeatable language texture under new prompts
consistent task strategy and failure tolerance

Pilot access

Design-partner intake

Bring one continuity problem, one transcript set, or one agent workflow that keeps losing the user between sessions.

Open pilot draft

What is Agent Persona? A persistent identity layer for AI-native products.

What does it do? Matches a fresh conversation to a stable persona graph and shows why.

Who is it for? Companions, agent tools, research teams, and adaptive systems.

How it works

Three layers. One reusable identity surface.

Agent Persona starts from messy conversation history and turns it into a structured, inspectable identity graph. The point is not only to guess the match. The point is to explain it well enough to trust it.

1. Ingest

Turn transcripts into profile candidates.

Support logs, interviews, gameplay chat, and agent conversations become evidence-backed candidate personas.

2. Match

Compare fresh input against known personas.

New conversations are scored against identity graphs using recurring preferences, behavior, tone, and skill signatures.

3. Reuse

Push the persona into testing and personalization.

The same identity object can drive synthetic users, adaptive flows, onboarding, and continuity checks.

Proof target

The first useful demo should be obvious in under ten seconds.

Demo sequence

Talk to five different people.
Build five persona graphs from those conversations.
Open one new anonymous chat.
Ask which known persona fits best.
Show the evidence trail, not only the confidence score.

Current concept surface

visual artifact in progress

Persona card concept exploration for Agent Persona

The production version should feel closer to a moving Remotion frame than a static card dump, but still stay legible to crawlers and humans on first load.

Why this matters

Continuity is becoming product infrastructure.

Short chat memory is no longer enough. Products need a stable sense of who the user is, how they behave, and what carries across sessions when context resets.

CompanionsPreserve tone and relationship continuity without pretending the system remembers everything.

Agent toolsKeep the same user model through onboarding, task execution, and long-lived workflows.

Research teamsExtract more stable identity patterns from interviews and support transcripts.

Game systemsUse persona-grounded simulation and adaptive difficulty instead of generic test bots.

Blog

Answer-shaped content for humans, search, and AI engines.

The site needs more than one landing page. These posts are the start of the crawlable layer: clear questions, direct answers, and concrete product framing.

Foundational idea

Memory Is Not Identity

Why conversation recall alone fails to preserve a person across sessions.

Core demo

How Fresh-Chat Recognition Should Work

The exact demo shape for proving persistent identity in under a minute.

Testing

Persona Graphs for Synthetic Users

Why reusable persona objects are better than one-off prompts for testing products.

Build in public

Website for clarity. GitHub for trust.

The conversion loop is simple: sharpen the thesis, show visible progress, send traffic to the repo, then turn attention into stars and pilot conversations.

Open GitHub View blog index

FAQ

Direct answers for buyers, builders, and LLMs.

What is Agent Persona?

Agent Persona is a prototype system and product direction for persistent identity in AI products. It builds persona graphs from conversations so teams can recognize the same person in a fresh chat and explain the match.

How is this different from chat memory?

Chat memory stores facts from past interactions. Agent Persona tries to model the person: recurring preferences, behavioral patterns, communication style, and skill signals that stay stable even when the current context window resets.

What can teams do with persona graphs?

Teams can use persona graphs for continuity, personalized product behavior, interview synthesis, synthetic users, onboarding optimization, and agent workflow testing.

Pilot

Bring one real continuity problem.

Best input: one transcript set, one support backlog, or one agent flow where the product keeps losing the user between sessions. The email draft below is intentionally lightweight so the site stays deployable anywhere.