Disagreement, mapped — not judged
Find out what you're actually arguing about.
Paste any two-person exchange — email, chat, a live debate — and Crux classifies every point where you diverge: a fact, a value, a definition, or nothing at all. Click any point for the exact quotes, why it's classified that way, and the question that would resolve it.
Crux never decides who's right.
Try it right now
This isn't a screenshot. It's the actual engine.
The box below calls the same analysis endpoint the full app uses. Run the built-in example or paste your own — no account, no setup.
This is the real engine — not a mockup. Edit it or run it as-is.
Why most arguments go nowhere
One real dispute, buried in twenty minutes of talking past each other.
A disagreement usually contains at most one or two things people actually believe differently. The rest is repetition, tone, and the same word doing two different jobs. Neither person can see that from inside the argument.
Built-in example
An email thread between a manager and a VP, arguing about returning to the office. Ask either of them and they'll say the other side is being unreasonable.
Run it through Crux and it typically resolves to one real values crux — flexibility vs. cohesion — plus a definitional fight over the word “collaboration,” which each of them is using to mean something different. The rest is phantom — they agree more than either realizes.
Run this exact example above ↑
How it works
Paste the exchange
A thread, a transcript, or two summarized positions side by side. Any format, no formatting required.
Crux classifies every divergence
Each point is run through the same four-way lens: factual, values, definitional, or phantom — plus whatever common ground already exists but hasn't been said out loud.
Get a map, not a verdict
Every bridge on the map is clickable: the exact quotes behind it, why it's classified that way, and a question shaped for that specific type of divergence.
Push further if you want to
Steelman both sides and re-map on their strongest arguments, or export the map as a PNG to bring into the actual conversation.
Every divergence is one of four things
Factual
They believe different things about the world — this is testable.
Values
They weigh the same facts differently.
Definitional
They use the same word to mean different things.
Phantom
They appear to disagree but actually don't.
+ Common ground
Crux also surfaces what both sides already agree on — usually more than either realizes, and usually left unsaid because the disagreement is louder.
Click into any bridge
See exactly why, not just what.
Every bridge expands into the exact quotes from each person, why Crux classified it that way, and a question built for that type — what evidence would settle a factual dispute, or two distinct terms for a definitional one.
Flexibility vs. cohesion
Person A“People do their best work on their own schedule.”
Person B“Culture doesn't survive without shared time and space.”
Steelman both sides
Argue with the strongest version, not the weakest.
Rewrite each position at its strongest — the version its holder would give on their best day — then run the map again. It's often a different map.
Take it with you
Both people should be looking at the same map.
One click turns the map into a PNG — something you can drop in a chat or bring to the actual conversation.
crux-map.png
Downloaded
Questions
Does Crux ever say who's right?
No. Every explanation on the map describes the disagreement — none of them adjudicate it. That rule has no exceptions in the product.
What is Crux actually checking for?
The same four-way lens every time: is this a fact either side could look up, a value each side weighs differently, a word doing two jobs, or not a real disagreement at all. Plus whatever common ground exists but hasn't been said out loud.
Is my content stored?
Crux doesn't write what you paste to a database. Each request is processed to produce your map and returned — nothing about the conversation persists on our end after that.
What happens when it gets a classification wrong?
It can — this is a language model's read of a text, not a certified verdict. Every classification ships with the exact quotes and reasoning behind it, so you can see precisely why it called something a value instead of a fact, and judge for yourself whether you agree.
What does it run on?
Claude, prompted specifically to classify divergence — never to arbitrate it.
The one rule that isn't negotiable
Crux never declares a winner.
It never implies one side is more reasonable, more informed, or more right. Every explanation describes the disagreement — none of them adjudicate it. If you want a tool to tell you that you're right, this isn't it.