Skip to content

Personas & the social map ​

A persona is a social map: Neuronz.ai remembers you and the people you work with — a colleague's role, their working style, what to avoid, the project they're driving — and how each of them relates to you. Your own traits are in view from the first message; a colleague's are pointed at only when a prompt concerns them.

What a persona holds ​

A subject is anyone Neuronz.ai keeps a persona for: you (created automatically) or a colleague (added as you tell your agent about someone). A subject can also be a group, a class of people like "clients" or "the team". A group carries traits exactly like a person ("clients expect formal French") and is labelled as a group wherever it appears.

Every subject has:

  • A name and aliases: the forms a prompt might use to refer to them (first name, nickname, handle), so a mention of any of them points to the right person.
  • Traits: short facts about that one subject, such as their role, skills, tone, what to avoid, their current project, or how they relate to you. A trait is a fact attached to its subject, so it avoids duplicates and keeps its history like any other fact.
  • A summary: a short, readable overview of who they are. It is for you: it shows on the dashboard and comes back from get_persona. Your agent is never handed the summary on a prompt; it gets the traits.

Notes about a colleague are private to the profile they were written in and are never visible from another profile.

How traits reach your agent ​

Your own traits. When a session opens, up to 25 of your traits ride the session-start briefing in full. After that, your agent is reminded of your top 5 traits every 20 prompts and right after the session's context is compacted. In between, no reminder is sent.

A colleague's traits. On every prompt, Neuronz.ai checks whether the prompt names a colleague by their name or one of their aliases. For each one it finds, your agent gets a short note instead of the traits: the name, how many traits are stored about them, and an instruction to fetch them with get_persona before writing to or about that person. Ask your agent to draft a message to Alice, and it is told that ten traits about Alice are stored and not yet in front of it; ask something unrelated, and Alice is not mentioned. At most 4 people (you included) are named for one prompt.

Contradictions ​

If two traits about the same subject contradict each other, they appear in the Conflicts tab of the Persona page in the dashboard. Both stay out of recall until you keep one, retire them all, or replace them with a correction. A warning icon beside Persona in the sidebar shows how many still need an answer.

Keeping the summary current ​

When your agent adds, edits or retires a trait, it gets back the subject's current summary and how much has changed since the summary was last rewritten, and it refreshes the summary in the same turn. The dashboard flags a summary that is due a refresh. You can also ask your agent to rewrite a summary, or set it yourself with update_persona_summary.

Using it ​

Tell your agent about yourself the way you always have: a stated fact becomes one of your traits. To record something about a colleague, name them when you say it ("remember that Alice prefers async updates"), or pass subject to fact_add.

The persona tools, available over MCP:

  • get_persona: with no name, your own profile; with a colleague's name, theirs.
  • list_persona_subjects: everyone on this profile.
  • add_persona_subject, update_persona_subject, delete_persona_subject: manage the list of people and groups.
  • update_persona_summary: rewrite a subject's summary (yours or a named colleague's).
text
> get_persona("Alice")

subject   Alice (alias: Al)
summary   Backend lead on the billing service. Prefers async updates over meetings.
traits    role: backend lead · tone: direct, prefers written summaries ·
          current project: billing-v2 migration

You can review and edit everyone, including yourself, on the Persona page of the dashboard.