NoFlattery

An AI council that argues, instead of agreeing with you.

Visit Website
June 19, 2026 I built an AI council that argues with me because every AI I asked just agreed

I almost shipped a product with the wrong name.

I'd picked something that sounded powerful. I asked ChatGPT - it loved it. Asked Claude - it nodded along. Nobody warned me about the trademark conflict, the wrong search intent, or the SEO fight I'd pick with the BBC.

That's when I realized the problem wasn't the name. It was the feedback loop. Most AI assistants are tuned to please, so they hide your blind spots instead of showing them.

So I built NoFlattery.

It's a local-first AI council. You put 2–4 agents in a room, give them a decision you're stuck on, and they argue. Each agent has a different bias - Analyst, Strategist, Humanist, Integrator — so they actually challenge each other, not just give you four separate answers. The output is a Decision Record: verdict, reasoning, main risk, what would change the call, and a next step.

Technically, it's a single-file SPA: React 19, Zustand, Dexie for IndexedDB, Vite + vite-plugin-singlefile. Your chats and API keys never leave your browser. BYOK. One-time price. No account, no subscription.

I built it solo because a tool about honesty shouldn't harvest your data or trap you in a subscription.

Has anyone else tried building with multi-agent disagreement? I'd love to hear what worked — and what didn't.

https://noflattery.com

1 Comment

  1. 1

    Common question - do you need multiple API keys?

    Nope. One key, all agents. Each agent has a different reasoning bias, not a different model. Works fine with just one OpenAI or Google key.

    Pro tip: mixing models across agents makes the debate sharper. Different models are trained on different data, so they genuinely disagree more. OpenRouter is the easiest way to do this — one key, tons of models, some free.

About

I got tired of asking AI for advice and getting a polite version of my own opinion back. Every model I tried nodded along, even when my idea was bad. So I built the opposite: a council of AI agents that disagree on purpo