Mediator.ai is a tool that applies Nash bargaining game theory and LLMs to facilitate fair cooperative negotiation, generating and scoring candidate agreements against both parties' stated needs until an optimal solution is found.
Eight years ago, my then-fiancée and I decided to get a prenup, so we hired a local mediator. The meetings were useful, but I felt there was no systematic process to produce a final agreement. So I started to think about this problem, and after a bit of research, I discovered the Nash bargaining solution.<p>Yet if John Nash had solved negotiation in the 1950s, why did it seem like nobody was using it today? The issue was that Nash's solution required that each party to the negotiation provide a "utility function", which could take a set of deal terms and produce a utility number. But even experts have trouble producing such functions for non-trivial negotiations.<p>A few years passed and LLMs appeared, and about a year ago I realized that while LLMs aren’t good at directly producing utility estimates, they are good at doing comparisons, and this can be used to estimate utilities of draft agreements.<p>This is the basis for Mediator.ai, which I soft-launched over the weekend. Be interviewed by an LLM to capture your preferences and then invite the other party or parties to do the same. These preferences are then used as the fitness function for a genetic algorithm to find an agreement all parties are likely to agree to.<p>An article with more technical detail: <a href="https://mediator.ai/blog/ai-negotiation-nash-bargaining/" rel="nofollow">https://mediator.ai/blog/ai-negotiation-nash-bargaining/</a>
# Mediator.ai - Cooperative negotiation is a solvable problem
Source: [https://mediator.ai/](https://mediator.ai/)
Daniel ran out of money six months in and took a delivery job to keep the lights on\. He still bakes every morning at 5am, but for the last 18 months, Maya has run everything else\. The staff\. The suppliers\. The books\. The Instagram that took them from 400 to 11,000 followers\.
An investor now wants 20% for $80k\. He won't wire the money until Maya and Daniel clarify who owns what\. Maya thinks 70/30 is fair, and she can point to four times the hours\. Daniel thinks a handshake is a handshake, and he can point to the delivery income that's covered their shared rent for eighteen months, and to the bakery profits he hasn't taken a cut of in over a year\. They can't agree\. They're about to lose the deal\.
Maya and Daniel each walked through their side with Mediator privately\. Then Mediator got to work: drafting candidate agreements, pitting them against each other, scoring each against both sides' needs, round after round, until no new draft could do better\.
What it surfaced was something neither Maya nor Daniel had proposed\. And something neither would walk away from:
### The split: 60/40, with a way back\.
Daniel can restore his 10% either by returning to full\-time for six months, or by forgoing $24k in distributions over two years\. It's not permanent, it's not punitive, and it ties ownership to what happens next, not to a painful audit of the past\.
*That's*what makes the deal work\. Daniel isn't being pushed out; he's being given a choice\. Maya isn't absorbing future resentment; she's acknowledging reality\.
The agreement also added a management salary so Maya's extra hours are paid in cash, not equity\. A clause waiving all claims on the first 18 months\. A shotgun buy\-sell if it ever falls apart\.
[Read the full agreement →](https://mediator.ai/examples/bakery-partnership-agreement)
Or see what each of them wrote privately, before Mediator ran:[Maya's statement](https://mediator.ai/examples/bakery-partnership-maya)·[Daniel's statement](https://mediator.ai/examples/bakery-partnership-daniel)
This paper introduces an automated mediator for human negotiation that uses a structured pipeline of LLM modules to conduct pre-mediation. In human-subject experiments, the system achieves preparation outcomes comparable to professional human mediators while reducing error in preference inference.
This paper introduces a framework using reinforcement learning from verifiable rewards to train large language models for strategic bargaining in multi-buyer markets, addressing private information and surplus extraction in concurrent negotiations.
This paper proposes a novel preference estimation method that integrates natural language information from LLMs into a structured Bayesian opponent modeling framework for multi-agent negotiation. The approach leverages LLMs to extract qualitative cues from utterances and convert them into probabilistic formats, demonstrating improved agreement rates and preference estimation accuracy on multi-party negotiation benchmarks.
This paper investigates formal mechanisms, such as Mediation, to maintain market stability among self-interested LLM agents (DeepSeek-V3) in a simulated marketplace, finding that Mediation enables recovery even under sustained adversarial attacks.
PACT introduces a head-to-head negotiation benchmark for LLMs using a 20-round buyer-seller bargaining game to test persuasion and adaptation. Top performers include GPT-5.5 and Opus 4.7, with ratings computed via Glicko-2 on an Elo-like scale.