signal/raiseTrack rounds
SeedAI · Israel

Plurai raises $13m seed round led by Team8 to test AI agents in simulation

The Israeli startup, founded by two former Nexar AI researchers, applies the simulation methods used for self-driving cars to enterprise AI agents.

s/rFunding deskIsrael · 3 min read

Plurai, an Israeli startup that tests AI agents in simulated environments before and after they go live, raised $13m in a seed round led by Team8. The company announced the round on 11 November 2026 as it came out of stealth.

U&I, Velocity and Mercer also invested, along with more than 20 angel investors. Calcalist named CyberArk founder Udi Mokady and Reah Miyara, a senior director at Google Cloud who previously led model evaluation at OpenAI, among the backers. Dr Eyal Carmi, chief technology officer of Microsoft 365 Copilot, has joined the company's advisory board, Calcalist and Ynet reported.

No valuation was disclosed. Plurai also made its platform generally available, with a free tier, after previously limiting it to selected partners and customers.

“AI systems fail when they hit challenges down the road, which no one predicted.”Dr. Ilan Kadar, co-founder and CEO, Plurai (as quoted by Calcalist)

What Plurai does

Plurai builds a digital twin of an AI agent and the environment it works in, using the customer's own knowledge base, tools and policies. It then sends thousands of simulated users at the agent in a sandbox, some of them deliberately hostile, to find problems such as policy breaches, invented answers, wrong actions, broken tool calls, failed tasks and leaks of sensitive data.

The same system generates fixes for the failures it finds and runs the agent through the simulations again to check that the fix worked without causing new problems. Plurai also uses the simulation data to train small language models for each agent. These act as monitors in production, checking every interaction at a fraction of the cost of large frontier models, so companies do not have to rely on reviewing a small sample.

Developers can use the platform from coding tools including Claude Code, Codex, GitHub Copilot and Cursor, and it can test agents built on Salesforce Agentforce.

What the company says it has achieved

Plurai says its design partners have cut agent development cycles from months to hours, reduced failures that reach real users by 82% and lowered the cost of running reliable agents by 97%. In joint work with Nvidia and Nebius for an enterprise customer, it said, a model it post-trained beat OpenAI and Anthropic models on accuracy at 40 and 44 times lower cost respectively. These figures come from the company and have not been independently checked.

Team8 said Plurai works with Fortune 500 companies and large financial firms, and that Gartner listed it as a representative vendor in its 2026 market guide for AI evaluation and observability platforms. Aviad Harell, a managing partner at Team8, told Ynet that simulation would become as basic a part of the agent stack as continuous integration and delivery is for software.

The founders and the plan

Plurai was founded in 2025 by chief executive Dr Ilan Kadar, who holds a PhD in computer science and machine learning, and chief technology officer Dr Elad Levi, who holds a PhD in mathematics. Both worked at Nexar, which develops technology for autonomous driving; Kadar was vice-president of AI and Levi an AI architect. TechCrunch reported that Kadar worked with Waymo to simulate and evaluate its driving software.

Kadar told TechCrunch that when AI agents arrived, the founders saw the same pattern they had seen in self-driving: a demo was easy to build, but a reliable and affordable product was not.

According to Calcalist, Plurai has about 12 employees in Israel and the United States and plans to double its headcount and expand its US operations by the end of 2026. The money will go into research and development, sales to large companies and growth in the US.

Company website: plurai.ai