[Coin-announce] New COIN-OR project: Frontier, multi-objective decision optimization through AI agents

[Cameron Afzal]

Hi COIN-OR community,


I'm pleased to announce Frontier, newly added to the COIN-OR repository: an
open-source toolkit for multi-objective decision optimization, driven in n
atural language through an AI agent.


Frontier exposes an optimization engine as an MCP (Model Context Protocol)
server with skills, so an AI assistant can translate a decision described
in business terms – which projects to fund, how to split a budget, whom to
source from – into a model, solve it, and explain tradeoffs.


Under the hood it maps the Pareto frontier with pymoo's NSGA-II/III, then
optionally certifies the finalists against an exact solver – HiGHS on CPU
or NVIDIA cuOpt on GPU – to a measured optimality gap. The heuristic
explores broadly; the exact solver proves what you would commit to as you
refine the formulation.


In spirit it is an interface project: where PuLP made solvers usable from
Python, Frontier aims to make multi-objective optimization usable through
agents – including the COIN-OR-hosted cuOpt and HiGHS it drives as exact
backends.



   - COIN-OR repository: https://github.com/coin-or/frontier
   - Development and issues: https://github.com/cafzal/frontier
   - License: Apache 2.0


Feedback and contributions are welcome via the development repository.


Best regards,

Cameron Afzal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/coin-announce/attachments/20260708/dd18b333/attachment-0001.html>