# Personas

Canonical page: `https://nanthai.tech/features/personas`

Tier: Pro

Create a Persona once, then use it as your chat participant, a private Advisor, a scheduled-task specialist, or part of a multi-model discussion.

Build specialised AI identities with a custom name, avatar, system prompt, model, temperature, and tool access. Reuse the same Persona across direct chat, private consultation, scheduled work, and multi-model discussions.

## How it works

1. **Create a persona** — Give your persona a name, avatar (emoji or image), and a system prompt that defines its personality, expertise, and behaviour.
2. **Configure its model and settings** — Choose which AI model the persona uses, set its temperature and token limits, and decide which tools (search, integrations, knowledge base) it can access.
3. **Use it anywhere** — Assign a Persona to a chat, consult it privately as an Advisor, use it in a scheduled task, or add it to a multi-model discussion. Its identity and settings follow it everywhere.

## What you can do

- **Custom name and avatar** — Give each persona a unique identity — a name, emoji, or uploaded image. They show up in chat with their own avatar.
- **System prompt personality** — Define exactly how the persona should behave — its tone, expertise, constraints, and priorities. Full control over the AI's personality.
- **Per-persona model and settings** — Each persona can use a different model and temperature. Your code reviewer runs at low temp for precision; your creative writer runs at high temp for imagination.
- **Tool access control** — Grant or restrict access to search, integrations, and knowledge base per persona. Your research assistant gets web search; your writing coach doesn't.
- **Multi-model persona comparison** — Use the same persona across multiple models to see how different AIs interpret the same personality and instructions.
- **Assign to scheduled tasks** — Each step of an automated pipeline can use a different persona. Chain specialised AI identities together for complex workflows.
- **Persona-specific memory** — Memories are available across personas, so each one knows your preferences and context while maintaining its own personality.
- **Private Persona Advisors** — Ask up to three Personas to review a turn privately. Their advice streams into collapsible cards and the main participant synthesises one final response. On multi-model text turns, all participants receive the same private Advisor notes; there is no per-participant assignment. Advisors can be used once or kept available for future turns, with optional internet access.

## Perfect for

- A code reviewer persona that uses a low temperature, has a security-focused system prompt, and only produces structured feedback in your preferred style.
- A writing coach that encourages creative exploration, suggests vocabulary improvements, and follows your chosen style guide.
- A research analyst that has web search enabled, access to your knowledge base, and produces structured reports with citations.
- A meeting assistant with calendar integration that can check availability, draft agendas, and summarise meeting notes.
- A product decision privately reviewed by a customer advocate, technical lead, and risk specialist before the main participant responds.

## Related features

- Memories (`/features/memories`)
- Knowledge Base (`/features/knowledge-base`)
- Automated Tasks (`/features/automated-tasks`)