Marcus: Car-Seller Negotiation Agent

What it is

A Linq agent that plays the seller of a used 2008 Camry: it texts like a person, anchors at the ask, concedes in shrinking steps, and handles several buyers texting at once as one conversation. Built as the adversarial sparring partner for the Closer negotiator agent.

The problem

Testing a negotiation agent needs a realistic, adversarial counterparty that behaves like a real seller.

How it works

1. A per-chat debounce window means one turn runs per burst of messages, so group chats get one reply and chats never block each other
2. Cross-chat awareness: before each reply the agent reads every conversation on the line and folds live offers (amounts, not identities) into the prompt; the identity leak is prevented at the data layer
3. Stateless: each turn re-derives buyers, offers and commitments from the Linq transcript, so a restart loses nothing
4. Runware powers both the negotiation brain and the generated car photos

Stack

Disciplines

For your business

Persona-driven SMS agents and simulated counterparties for stress-testing a client's sales or negotiation bots.