Plurai raises $13 million and launches AI agent simulation platform
The AMW Read
The funding and launch add detail to an already reported event, with potential impact focused on agent testing and reliability.
Plurai raises $13 million and launches AI agent simulation platform
New York-based Plurai has raised $13 million and launched a platform that tests AI agents against synthetic, multi-turn scenarios. Its approach aims to find failures before agents interact with customers. The company also offers a process it calls vibe-training, which turns plain-language rules into specialized small language models for enforcing policies. Plurai has released IntellAgent, an open-source framework for simulating and evaluating conversational AI.
The launch adds a specialist to the market for tools that evaluate and control agent behavior. Reliability remains a barrier to using agents for complex, multi-step work without close human supervision. Simulation can expose failures before deployment, but its value depends on whether synthetic scenarios reflect the situations agents encounter in practice. Plurai also reports work with NVIDIA's Nemotron models and NIM deployment software.
For builders evaluating the platform, the practical test is whether it finds consequential failures in their own workflows and whether its guardrails work within their latency and cost limits. Plurai claims more than 43% fewer evaluation failures than GPT-5-mini, eight times lower cost, and up to 15 times greater edge-case coverage. These are company claims, and the article does not establish independent validation. Named customers and production deployments would provide stronger evidence of demand.
