⚠ DisclosureAn AI product→ Filed 04 · 2026 / Issue №01read time ~ 3 min

I'm not a pessimist.
I'm an optimist with
receipts.

Look — we get it. The internet is on fire. Every other tab is a chatbot named “Skyler” hallucinating at your customers. You don't want another one of these on your site[1].

Neither do we, really. So we built the version we'd hate the least.

See the receipts → ten cited answersRead what it won't do

[1] If your position is “no chatbot, ever, full stop” — totally fair. Close this tab. We won't email you. We don't even have your email.

§01   The 10 promisesUpdated April · 2026

Things we swore not to do, in writing.

An AI product that doesn't market itself with the word “revolutionary” is a refreshing change. So is one that publishes its constraints upfront. We did both. You're welcome.

№01
We will not

Hallucinate.

If FIM doesn't know, it says so. It will not invent a price, a listing, an opening hour, or a Tuesday-only special. Every fact comes from a page on your site. With a citation. Receipts attached.

№02
We will not

Pretend to be human.

The greeting reads "Hi — I'm a manager." Not "Hi, I'm Sara :)". Your visitors are adults. They figured out it's a bot the moment they saw a chat bubble.

№03
We will not

Talk like a hype reel.

No "transform your business," no "unlock potential," no "synergy." We charge $49. You'll get more leads. That's the whole pitch.

№04
We will not

Train on your data.

Conversations stay in your tenant. Nothing is sent to OpenAI for training. Nothing is sent to us for training. We don't even have a "training" pipeline. We have one prompt and a vector database.

№05
We will

Show our homework.

Every answer the bot gives links to the page it pulled the fact from. You can audit it. Your customer can audit it. Open-source-grade transparency on a closed-source product.

№06
We will

Refuse, often, on purpose.

"Is this product good for my dog's allergy?" — out of scope. The bot will say so. We'd rather lose a conversation than give a confident wrong answer that lands you in court.

№07
We will not

Auto-renew.

Trial ends, the bot stops. You get an email. If you want to keep it, you pay. If you don't, you don't. No "we kept charging your card for 11 months because you forgot."

№08
We will

Hand you the off switch.

One toggle in the dashboard. Bubble disappears in < 5 seconds across every page on your site. No support ticket, no "let me check with my manager."

№09
We will not

Add 47 features no one asked for.

Read the site, answer questions, capture warm leads. That's the product. We are not adding "AI sentiment analysis" because Hacker News told us to.

№10
We will

Pick up the phone.

Not the bot. Us. The two humans who built this. If something breaks at 11pm and you have a viewing tomorrow, you have our numbers. Try that with the big platforms.

§02   Receipts, exhibit BRecorded transcripts

Two answers. One product. Spot the difference.

Same question. Real real-estate site. Left: a generic chatbot guessing. Right: FIM, which read the listings.

The other guys№ 0118 · 22:14
Visitor →
“Do you have apartments in Kotor under €200,000?”
Generic chatbot ←
“Yes! We have many beautiful apartments in Kotor in your budget. Could you please tell me your name and email and an agent will be in touch to discuss options?”
⚠ sourcenone. it made that up. it doesn't know if you have apartments in Kotor. it just wanted the email.
↓ vs. ↓
FIM№ 0119 · 22:14
Visitor →
“Do you have apartments in Kotor under €200,000?”
FIM ←
Yes — three match right now. The closest is in the Harbor District, 62 m², €189,000, balcony facing the bay. Want me to set up a viewing or send the floor plan?
→ sourceacmeproperties.example/listings/harbor-62m2-sea-view — chunk #14 of 22
Made-up facts
0
Across 12,400 conversations on production tenants since launch. We grep for “I'm not sure but…” the way other companies grep for revenue.
“Great question!” detected
0
Banned at the system-prompt level. If you catch one, we'll buy you a coffee.
Citations per answer
≥ 1
Or the bot refuses to answer. There is no third option.
§03   Termination clauseArticle XII · always available

One big red button.
That is the contract.

You don't trust AI. Smart. Neither do we, all the way.

So we built the kill switch first and the product around it. One toggle. One click. Bubble vanishes from every page on your site within five seconds. No call, no email, no “are you sure” funnel.

This is not bravery. This is just how a tool should work. If you can't turn off your tools, they aren't tools — they're tenants. We'd rather be a tool.

§04   The billCost transparency, the unsexy kind

One price. One.
Not “starting from.”

“Starter / Pro / Enterprise” pricing exists to confuse you. We can't be bothered. $49. Done.

no card · 14 day trial

$49./ mo

That's the price. There's no second tier. There's no asterisk. We'll add one tier the day we genuinely need to, and we'll tell you why.

billed monthly · cancel anytime · trial does not auto-convert · we'll
email you 3 days before it ends and that's the only marketing email we send

Included.

  • Unlimited conversations — actually unlimited, no fine print
  • Re-crawl your whole site, weekly
  • Lead emails the second they qualify
  • Brand colors, your voice, your assistant name
  • SR / HR / EN out of the box · 80+ more languages
  • EU hosting · GDPR-clean · no US subprocessors

Not included.

  • “AI agent” anything
  • A sales call
  • A 47-page slide deck
  • The phrase “transform your business”
Start the trial — no card →
§05   Frequently honest answersQ&A · skeptic edition

The uncomfortable questions.

Q.01
“Aren't you just another AI grift?”
Honest answer: maybe. Time will tell. What we can prove today: every answer cites a page, the bot refuses out-of-scope questions, and you can rip it off your site in one click. That's three more guarantees than most of our competitors offer combined. Try it for 14 days, then judge.
Q.02
“What happens when the AI hallucinates and I get sued?”
Possible, but we've stacked the deck against it: retrieval-grounded answers, refusal on missing context, vertical-specific guardrails (medical > retail). We log every conversation; you have a full audit trail. We can't legally indemnify you against an LLM, but neither can anyone else honestly — be wary of those who claim they can.
Q.03
“What stops you from training on my customer chats?”
Two things. One: our data-processing agreement, in writing, EU-jurisdiction. Two: we literally don't have a training pipeline. Anthropic's API is set to no-train. Adding training would be a 6-month engineering project we have zero plans to start.
Q.04
“What if you get acquired by a creepy ad-tech company?”
Then you cancel and export your data. Your DPA stays in force until renewal; the new owner can't quietly change the rules without your re-consent. We aren't planning to sell. But a planning-document is not a guarantee, so the export endpoint exists and always will.
Q.05
“My customers hate chatbots. Why would I add one?”
Your customers hate useless chatbots — the kind that loop them through “Did this answer your question?” until they rage-close the tab. A bot that actually answers (with citations) tests as a feature, not a tax. Or maybe yours hate all bots. Then don't add one. We're not for everyone, and we'd rather you find that out in a 14-day trial than via a refund request.
Q.06
“What about the environment? All these tokens.”
Real concern, often dishonestly answered. Honest answer: a bot conversation runs on a few hundred tokens of inference, which is roughly the energy of a Google search. Not zero. Not catastrophe. We use Anthropic for chat (efficient) and OpenAI for embeddings (one-time per crawl). If carbon footprint is your dealbreaker, our footprint per $49 customer per year is < one transatlantic flight.

So. Convinced?
Doesn't matter.

We charge nothing for 14 days. You don't even need a credit card. The worst case is you spend ten minutes pasting a script tag, watch the bot answer your own questions about your own site, find a flaw you didn't expect, and email us about it. Honestly that would be useful for us too.

Start the trial — no card →…or re-read the 10 promises first