CRV and Lightspeed Venture Partners led a funding round for Raindrop to scale its safety layer designed to catch hallucinations and tool call errors in AI agents. The Series A brought the company's total funding to $50 million and included participation from researchers at OpenAI, Anthropic, and Thinky Machines. The platform detects failures in production that traditional evaluation methods miss, serving a client base that includes Fortune 100 firms.

The company also introduced Raindrop Simulations, which automatically creates mocked services and databases to test agent changes on every pull request. This tool enables developers to replay thousands of historical traces on new harnesses and is currently in early access with a general release scheduled for next month. Current customers of Raindrop's software include Vercel, Clay, Framer, and Speak.