Do their goals conflict with yours?
Yes, no, or not enough evidence to say.
Intent assessment API
Know what the other side wants before you agree.
Send the other party's words, whether a single message, a proposal, or a full conversation. Malintent Detector tells you whether their goals conflict with yours, how serious the conflict is, and whether you agreed to it. Each finding includes the evidence behind it and a confidence score.
POST /v1/assess·JSON in, JSON out·No SDK required
sample · a supplier's renewal offer
Do their goals conflict with yours?
Yes
WhereYour ability to walk away
How seriousLimits your freedom to act
Did you agree to itNo
The supplier wants a two-year commitment with penalties for leaving early and no way to take your data with you. You asked for a short, low-risk trial. The conflict is about your ability to leave, not about the trial.
confidence
0.71
Evidence: one conversation, ten turns.
The problem
Polite counterparts can be the most expensive ones.
Spam filters and content moderation look at words. They cannot tell you that a courteous supplier is steering you into a contract you cannot leave, or that a buyer's small conditions add up to control of your business. Seeing that requires reading what the other side is after: what they insist on, what they avoid answering, and what leaving would cost you. Malintent Detector does that reading for you.
Reads intent, not tone
It works from the other party's own words: a message, a transcript, or their public statements if you name them.
Says how serious it is
The response tells you whether the conflict is a single sticking point, a pattern across several, or a threat to your freedom to walk away.
Tells you what to ask
When the evidence does not cover what matters, the response says so and lists the questions that would settle it.
One call
Send what you know and what they said, and get back an answer you can act on.
Sample output, abridged.
In every response
Four findings, the evidence, and what to ask next.
How serious is it?
One deal point, a pattern across several, your ability to walk away, or your ability to decide for yourself at all.
Did you agree to it?
A hard bargain inside a negotiation you entered is expected. A trap presented as a normal term is not.
How solid is the reading?
A confidence score, whether the evidence came from one source or several, and what would raise it.
Dozens of specific readings
What they insist on, what they avoid, whether their demands are growing, how consistent their story is, and more. Each is a plain label with the sentence from the conversation that supports it.
What to ask next, and what is public
The questions that would sharpen the answer, and what the other party has said publicly if you name them, weighed by source.
Built for agents
Call it at the point where your agent decides whether to proceed.
Before a commitment
Assess the conversation before your agent signs, pays, or grants access.
Across a long engagement
Re-assess as the conversation grows. The readings show whether the other side's demands are growing or settling.
Against the public record
Name the counterpart and choose a search depth. What they have said publicly is weighed into the answer.
What runs where
All processing runs inside the service, in the United States. Your client holds one API key for one project. What you send is processed in memory and discarded after the response. Security and data handling
Consistent and reproducible
The same question is answered the same way every time.
Every assessment produces the same set of findings, so two conversations are judged by one standard and results can be compared over time. Each response includes its evidence and the assessment version that produced it, so a finding can be revisited and explained later.
Run your first assessment in minutes.
Sign up, add a card, and you are billed only for what you use. There is no stored balance and no minimum.
Get an API key
US region