Plurai Introduces Plurai Simulation: The Self-Improving Layer for AI Agents
Plurai, building the self-improvement layer for AI agents, led by Dr. Ilan Kadar and Dr. Elad Levi, has introduced Plurai Simulation, the self-improving layer for AI agents. It’s free to start for every AI engineer, and it has also announced $13M in seed funding led by Team8, with U&I, Velocity, and Mercer, and more than 20 angels and industry leaders. They include CyberArk founder Udi Mokady and Reah Miyara, Senior Director of Google Models at Google Cloud and former model evaluation lead at OpenAI. Dr. Eyal Carmi, CTO of Microsoft 365 Copilot, has joined the advisory board.
The founders spent years developing AI for autonomous driving, including work with companies like Waymo, where simulation is critical to identifying rare failures. They saw a similar challenge emerging in AI agents: promising demos often break under real-world conditions, while manual testing and expensive models make continuous improvement difficult.
Plurai was founded to address this through a recursive self-improvement layer for AI agents. Its platform creates digital twins of agents and their operating environments, then simulates realistic users, scenarios, and edge cases to identify failures before deployment. It also explains failures, generates fixes, and post-trains models to improve performance while reducing operating costs. In production, new failures feed back into simulation, creating a continuous cycle of testing, fixing and improvement.
Plurai tested its platform with design partners across AI-native and regulated industries, including Fortune 500 companies and major financial institutions. The company reports reducing agent development cycles from months to hours, cutting failures reaching users by 82% and operating costs by 97%. In collaboration with NVIDIA and Nebius, Plurai-post-trained models also outperformed OpenAI and Anthropic frontier models on accuracy at 40 and 44 times lower cost, respectively.
Plurai is now making its simulation platform available for free to agent developers through tools such as Claude Code, Codex, and GitHub Copilot, while also supporting enterprise platforms including Salesforce Agentforce.
The founders thanked their design partners, investors at Team8, and the Plurai team for supporting the company’s journey. Drawing a parallel with autonomous driving, the company believes AI agents can earn trust through continuous simulation and testing. Plurai is inviting developers to try its simulation platform for free and share their feedback.
