Describe a deal you're walking into — get your leverage, their likely moves, and how to respond
A salary talk, a vendor contract, a business sale — anything where you'll be sitting across from someone trying to get a different outcome than you.
Weighing your leverage, predicting their opening, working out your counters...
I noticed I'd rehearse important conversations in my head before they happened — a salary talk, a vendor negotiation, a tough ask — but I'd only ever think through my side. I rarely forced myself to seriously predict what the other person would actually say back, or have a real answer ready when they pushed. I wanted a tool that makes me do that work in advance: name my leverage honestly, admit my weak points instead of ignoring them, and rehearse the pushback before it happens for real.
What you get: a read on your leverage and your weak points, a prediction of the other side's likely opening position, a role-play of the specific pushback they might give and how to respond to each one, a suggested opening move, a walk-away sanity check, and a short list of questions worth asking them.
This is AI-only from the first click, like my Meeting Notes tool — there's no rule-based scoring underneath it, because "predict what a specific person will say in a specific negotiation" isn't something a formula can do. Your scenario, goal, walk-away point, and anything you know about the other side go into a single prompt that asks Claude to do several distinct jobs in one pass: assess your leverage and weaknesses honestly (including weaknesses you didn't explicitly mention if they're implied by the scenario), predict the other side's likely opening position, generate 3-4 specific moves they might make with a coached response to each, suggest a concrete opening move, sanity-check your stated walk-away point, and suggest questions that would surface information you need.
Claude replies in a strict JSON format that my code reads and lays out on the page. Your API key is typed into a one-time popup, used for that single request, and kept only in your browser's memory for the session — never saved or sent anywhere else.
This is my third live Claude API integration, and the first one that asks Claude to role-play another person rather than just organize or analyze information I already gave it. It's a step up in reasoning difficulty: instead of extracting structure from a meeting transcript or scoring a business against fixed criteria, Claude has to model a person it's never met and predict how they'll behave — then coach me on responding to a version of the conversation that hasn't happened yet.