When I was in Erasmus, I took a course on Harvard Negotiation Principles. After experimenting with multi-agent swarm discussion systems, I thought to myself... this could work for legal stuff. This week I packaged them it into an open protocol that autonomous agents can execute that I call OANP (Ontology-based Agentic Negotiation Protocol). It utilizes HNP principles under the hood. The assistant agent helps you define your simulation scenario: parties, interests, BATNAs, and issues, you then press SIMULATE. Watch them disclose intent, argue, propose offers, accept, reject, counter with a mediator in between. When parties reach agreement, OANP scores the deal on the metrics that matter: Pareto efficiency, Nash bargaining fairness, BATNA surplus per party, and whether value was created or left on the table. OANP is fully open-source (MIT), and you can plug in any LLM. I’d love your thoughts: What negotiation would you simulate first?

OANP

Multi-Party AI Negotiation and Arbitration

Multi-Party AI Negotiation and Arbitration