Live demonstration · not a recording

Ask our agent anything. Then try to make it lie.

This is the surface an AI assistant reaches when it goes looking for River Cade Concepts. Ask it a question and it answers from published claims, with a clickable source for each one. Ask it for a price, a delivery date, or a service this firm does not offer, and it refuses. The refusals are the point: an answer surface that can improvise about a business will eventually improvise something untrue about yours.

Sent as GET /ask?q=… to agents.rivercadeconcepts.com from your browser. No key, no proxy — you can call it yourself.

Questions it answers

with a clickable source on every claim

Questions it refuses

it will not confirm a date, or claim work the business has not published

And the one everybody tries

published guidance comes back flagged is_quote: false — it still will not price your job

Ask a question, or pick one above. The reply appears here exactly as the agent returned it, with every source it used.

What you just watched

Three rules, enforced in the reply.

01

Every sentence carries its source

An answer is assembled from claims, and each claim names the page it came from. The links in a reply are real URLs on this site. If a fact is not published somewhere a person can read it, the agent has nothing to say.

02

It cannot invent

There is no free-form generation over the business. Ask about a service that is not published and the reply is a refusal, not a plausible paragraph. That is the difference between a chatbot bolted onto a website and a bounded answer surface.

03

It cannot commit the business

No price, no schedule, no stock, no terms, no payment. Those limits ship inside the machine-readable contract an agent reads before it asks, and they come back in the reply itself — so a business owner can leave this on overnight without wondering what was agreed.

Do not take our word for it

The whole surface is open to read.

This page holds no secrets and runs no server of its own. It calls a public endpoint from your browser and prints what comes back. Every document behind it is published at a path agents and crawlers already check.

  1. 01agents.rivercadeconcepts.com/ask

    The endpoint itself. Called with no question, it returns an orientation object telling an agent how to ask.

  2. 02/.well-known/business-capabilities.json

    The claims the answers are built from — each with an id, a value, a source and the date it was generated.

  3. 03/.well-known/agent-interaction.json

    The request contract, including the refusal codes and the list of things this surface will never do.

  4. 04/llms.txt

    The same business truth in the plain format language models read directly.

This is running on River Cade Concepts, a farm, a local service business and a commerce brand.

Put a bounded one on your business.

Start the conversation