Plurai creates a digital twin of a company's electronic systems, including the knowledge base and the tools an agent can reach in production. It maps the possible agent flows and uses that graph to evaluate the model and fix edge cases before deployment. The platform is designed to work alongside tools such as Claude Code, Codex, Copilot and Cursor.
The company says its main product is using the data gathered in these simulations to post-train small, open-weight models for individual customer needs. Tests run with Nvidia and Nebius showed its models outperforming mid-tier frontier models at lower cost, according to co-founder Ilan Kadar, who said the approach could cut cost by one or two orders of magnitude.
The seed round was announced on Wednesday and includes participation from U&I, Velocity and Mercer. Google DeepMind researcher Reah Miyara backed the company as an angel investor.
Why it matters if you build agents
Developers building agents that book travel, query several databases or handle long support chats can test failure cases against a copy of their own systems before going live.
The approach points toward replacing large frontier models with smaller, task-specific models for narrow workflows, which could lower inference costs if the claimed results hold up.
Get stories like this every weekday
One short email with the day's agent tooling news.