Plurai Simulation: A simulated world that makes your agent improve on every run
On November 11, 2026, Plurai, a New York startup, announced $13 million in seed funding led by Team8 and general availability for its simulation platform, which had been limited to select partners and customers. The product ships as a plugin for Claude Code, OpenAI Codex and GitHub Copilot. Pointed at an agent's code, it builds a surrounding test world with mocked tools that share one state, simulated users with goals, and evals that score what actually happened.
The system models the agent's environment from a customer's knowledge base, tools and policies, then runs thousands of realistic and adversarial users through a sandbox to expose hallucinations, policy breaches and tool failures. Each failure comes with a suggested fix that is checked against a regression set before release. Simulation data also trains small language models for each agent, which can run the agent or monitor it in production, and new production failures feed back into the simulation.
Plurai says design partners cut development cycles from months to hours, reduced failures reaching real users by 82% and lowered the cost of running reliable agents by 97%. Dealroom notes those figures come from early customer work and await independent validation. Eyal Carmi, CTO of Microsoft 365 Copilot, has joined the advisory board.
Why it matters
For teams building agents, the pitch moves testing from a fixed set of known cases toward generated scenarios that cover rare and adversarial paths before customers reach them. For enterprises that cannot send real customer data to outside testers because of privacy and compliance rules, synthetic environments offer a route to evaluation. The cost claim matters too: running small models for production checks instead of calling a frontier model on every interaction changes unit economics, if the numbers hold.
For investors, the round is a bet that evaluation and simulation become core infrastructure rather than add-ons. Dealroom reports that agent evaluation and observability firms raised $272 million through September 2026, up from $166 million a year earlier. Founders selling into this space should expect buyers to ask for independent benchmarks, and should test claims like the 97% cost reduction against their own workloads.
Key points
- 1Plurai raised $13 million in seed funding led by Team8 and opened its platform on November 11, 2026.
- 2The simulation runs thousands of realistic and adversarial users against an agent inside a sandbox.
- 3Design partners report 82% fewer failures reaching real users and 97% lower costs, figures not yet independently validated.
- 4The product ships as a plugin for Claude Code, OpenAI Codex and GitHub Copilot, with a free tier.
- 5
