Private AI deliberation rooms. Convene a small council of personas to think through a dilemma that doesn't fit a one-shot answer.

Like a few sharp friends arguing your dilemma out loud — none of them in a hurry to agree.

  1. 01

    Set a table.

    Type the dilemma in your own words. No prompt engineering.

  2. 02

    The council speaks.

    Personas take turns. You watch the thinking happen, token by token.

  3. 03

    A synthesis, not a verdict.

    What was said, woven together. The decision stays yours.

Try the demo

A short, three-persona deliberation pinned to a small model. Auto-deletes within 24 hours; anyone with the link can read it.

Self-host in 15 minutes

Your dilemmas, your API keys, your machine. Bring your own Anthropic / OpenAI / OpenRouter key — or run fully local with Ollama. Single SQLite file, no analytics, no telemetry.

Self-host guide →

What it isn't

  • Not a decision-maker. The council won't tell you what to do.
  • Not balanced. Opinionated councils tilt on purpose — you bring the counterweight from the rest of your life.
  • Not an advisor. No expertise theater, no confidence performance, no "as an AI…".

What gets sent where

  • Your dilemma text → this server, then forwarded once to the LLM provider configured for the demo (Anthropic, by default)
  • The deliberation is stored in a SQLite file on this server until it auto-deletes
  • No analytics, no tracking pixels, no third-party scripts
  • For private use, self-host — nothing leaves your machine except the LLM API call. Point Ahwa at a local Ollama install and even that stays on your hardware.