Tuesday, November 17, 2026Tue, Nov 17 Updated

Mergers, acquisitions and startup deals, and what they mean.

Plurai Raises $13M Seed Round Led by Team8 for AI Agent Testing

Plurai raised $13M in seed funding led by Team8 and opened its AI agent testing platform to all developers, with a free tier.

A woman using a laptop navigating a contemporary data center with mirrored servers.
Photo: Christina Morillo / Pexels

Plurai, a New York company that tests AI agents before they reach customers, announced $13M in seed funding on Nov. 11, according to a PR Newswire release. Team8 led the round, and the company also made its platform available to all developers.

Until now the product was limited to select partners and customers. It now has a free tier, so any team building an agent can try it.

Who put money in

Besides Team8, the round included U&I, Velocity and Mercer, the company said. More than 20 angel investors and industry leaders also took part. One of them is Udi Mokady, the founder of CyberArk. The release did not state a valuation.

Dr. Eyal Carmi, who is CTO of Microsoft 365 Copilot, has joined Plurai’s advisory board. Reah Miyara, a Senior Director of Google Models at Google Cloud and a former model evaluation lead at OpenAI, is an angel investor in the company.

How the platform works

Plurai builds a digital copy of a customer’s agent and the environment around it, using the customer’s own knowledge base, tools and policies. It then sends thousands of realistic and hostile simulated users at the agent in a sandbox. The goal is to surface policy violations, hallucinations, wrong actions, broken tool calls, failed tasks and leaks of sensitive data.

The same simulation data is used to fix the agent. Plurai generates fixes for the failures it finds and trains small language models, or SLMs, tailored to each agent. A small language model is a compact AI model built for a narrow set of tasks, as opposed to a large general-purpose one. In production, these models can take over from frontier models in running the agent, and they also check each action against company policy. Failures they catch go back into the simulation.

Developers can start from tools such as Claude Code, Codex and GitHub Copilot. Teams can also simulate and improve agents built on enterprise platforms such as Salesforce Agentforce.

The results Plurai reports

All performance figures come from the company’s own announcement. Plurai said design partners saw development cycles fall from months to hours, failures reaching real users drop by 82% and the cost of running reliable agents fall by 97%.

In joint work with NVIDIA and Nebius for a global enterprise customer, models that Plurai post-trained beat OpenAI and Anthropic frontier models on accuracy, the company said. The cost was 40x lower than OpenAI’s and 44x lower than Anthropic’s, respectively. The customer was not named.

Plurai also cited VentureBeat Pulse Research, which found that half of enterprises have deployed an AI agent or LLM feature that passed internal evaluations and still caused a customer-facing failure in production. Gartner named Plurai a Representative Vendor in its 2026 Market Guide for AI Evaluation and Observability Platforms. In that category, Gartner projects 60% of software engineering teams will be adopters by 2028, up from 18% in 2025, according to the release.

The founders

Dr. Ilan Kadar is CEO and Dr. Elad Levi is CTO. Both worked at Nexar, where Kadar was VP of AI and Levi was an AI architect. There they helped build real-time systems for autonomous driving companies such as Waymo, which use simulation to validate self-driving software before it goes on the road.

Plurai said its research includes BARRED, a method for generating synthetic training data for custom small language models, which was accepted at ICML 2026. Kadar said the company applies the simulation approach from autonomous driving to AI agents.

Why it matters for companies running AI agents

Businesses that put AI agents in front of customers face a testing problem. According to Plurai, teams spend months checking agents by hand and still cannot foresee every scenario. Using frontier models to run and grade every live interaction is also too slow and too expensive, the company said. Plurai’s pitch is a cheaper, always-on layer of checking that runs on smaller models.

Plurai said it already works with enterprises in regulated industries, including Fortune 500 and major financial companies. Mokady, an angel investor, framed the appeal in a short remark: "AI is probabilistic. Enterprises are not."

The announcement did not say how many customers pay for the platform, what the paid tiers cost or when the company expects to raise again. Readers evaluating the claims should note that the performance numbers are the company’s own and were not independently verified in the release.

Source: PR Newswire