Plurai raises $13M to stress-test AI agents with simulations

Plurai, a startup founded by two veterans of autonomous driving company Nexar, said it raised $13 million in seed funding. Team8 led the round, with U&I, Velocity and Mercer participating. The company applies simulation with synthetic data, a method used to prepare self-driving systems for rare situations, to AI agents.

The pitch responds to a reliability gap. TechCrunch notes that benchmarks show agents often need close supervision. In one simulated corporate test by EON, the best model scored 76.8% accuracy. Plurai says agents can ignore spending policies when pressed, struggle to combine multiple datasets, and lose their way in long customer conversations.

Plurai builds a digital twin of a customer's systems, including the knowledge base and tools an agent would use in production. It maps possible agent flows and uses that map to run evaluations and fix edge cases. It works with tools such as Claude Code, Codex, Copilot and Cursor.

For marketers and content teams, this matters wherever agents touch customer-facing work, such as support bots or workflow automation. Co-founder Ilan Kadar says the larger opportunity is using the simulation data to post-train small open-weight models for a specific customer. He claims this can match or beat frontier model quality at one to two orders of magnitude lower cost. Tests with Nvidia and Nebius, per the report, showed its models outperforming midtier frontier models at lower cost.

The claims come from the company and its backers, so teams should look for independent results. Watch whether Plurai publishes benchmarks on real customer deployments, and how its tailored models hold up against the frontier models they are compared with.

Sources

This brief is our own summary. Read the full reporting at the source.

• TechCrunch: Plurai wants to help AI agents work better using the same approach that helped get self-driving cars to roads