Plurai Raises $13M Seed and Opens AI Agent Simulation Platform

Plurai announced the funding and launch on November 11, 2026, disclosing $13 million in seed funding and general availability for its AI-agent simulation platform. Team8 led the round as the company introduced a free tier for a product previously limited to select partners and customers.

U&I, Velocity and Mercer also participated, joined by more than 20 angel investors and industry leaders. They included CyberArk founder Udi Mokady and Reah Miyara, senior director of Google Models at Google Cloud and a former model-evaluation lead at OpenAI. Microsoft 365 Copilot CTO Eyal Carmi joined Plurai’s advisory board.

Team8 managing partner Aviad Harell said Plurai addresses the infrastructure needed to build trust before agents reach customers and keep improving them in production, adding: “Simulation will become as fundamental to the agent stack as CI/CD is to software.”

How the Simulation Platform Works

Plurai builds a digital twin, or software replica, of an agent and its operating environment from the customer’s knowledge base, tools and policies. It then subjects the agent to realistic and adversarial simulated users in a sandbox, seeking policy violations, hallucinations, incorrect actions, broken tool calls, failed tasks and sensitive-data leaks before customers encounter them.

The product page describes isolated sandboxes in which services, documents and world state can be mocked and reset between runs, while production traces can inform later scenarios. Plurai traces failures to evidence and exact code locations, generates fixes and validates them against a regression set to check that one repair does not create another failure.

Simulation data also supplies examples for post-training, the additional training of small language models for narrow tasks. Those models can run the agent or act as continuous evaluators and guardrails, while newly detected production failures feed back into simulation. Developers can start from Claude Code, Codex or GitHub Copilot, and teams can use the platform with agents built on Salesforce Agentforce.

Company-Reported Results

Plurai reports that design partners cut agent development cycles from months to hours, reduced failures reaching real users by 82% and lowered the cost of running reliable agents by 97%. The results came from a select group across AI-native and highly regulated industries, including Fortune 500 and major financial companies, before the platform opened to a broader user base.

An August 15, 2026 customer case study provides a detailed look at one deployment. Plurai said an unnamed U.S. enterprise IT managed-services provider built its first custom model in 20 minutes and more than 100 within two weeks. For the portal-routing decision examined in the study, the custom model resolved 89% of requests, compared with 69% for the existing Claude Haiku setup without reasoning, while latency fell from 1.0 second to 0.15 seconds and inference cost from $2.87 to $0.076 per 1,000 calls.

A separate October 21, 2026 case study describes work with NVIDIA and Nebius for a global enterprise customer. Plurai reports that it post-trained NVIDIA Nemotron 3.5 Lightning using a training set of 3,722 conversations labeled by a frontier teacher model; the resulting model received a quality score of 88.8 out of 100 on real production chats, against 78.4 for Claude Haiku-5.5 and 88.6 for GPT-6 Luna.

Autonomous-Driving Roots and Synthetic Data

Chief executive Ilan Kadar and chief technology officer Elad Levi founded Plurai after working at Nexar, where Kadar was vice president of AI and Levi an AI architect. Kadar compared Plurai’s approach with autonomous-vehicle validation, citing Waymo’s more than 20 billion simulated miles, which he put at roughly 100 times its real-world driving.

BARRED, a preprint submitted on April 28, 2026 by Arnon Mazza and Levi, supplies the synthetic-data method behind that pipeline. It decomposes a domain into dimensions and uses multi-agent debate to verify labels, requiring only a task description and a small set of unlabeled examples. Plurai says the method was accepted at ICML 2026.

The seed round and free tier move Plurai from selected design partnerships to a broader self-service test of that approach. The central question for its next stage is whether the reliability and cost gains reported in those deployments hold as more teams adopt the full loop of simulation, automated fixes and specialized models.