Raindrop has raised a Series A led by CRV, bringing its total funding to $50 million as the company expands technology designed to detect failures in autonomous AI agents. Existing investors Lightspeed Venture Partners and Y Combinator participated alongside lead researchers from OpenAI, Anthropic and Thinking Machines. Raindrop is already being used by customers including Vercel, Framer, Clay and Fortune 100 enterprises.
The financing will support Raindrop’s anomaly-detection research, enterprise expansion and development of Simulations, a new product intended to identify unexpected AI agent behavior before changes reach production.
Traditional AI evaluations generally depend on predetermined test cases and therefore tend to measure problems engineers already anticipate.
Raindrop’s Simulations product instead replays actual production traffic and existing test cases against proposed changes to an agent’s underlying system. Its anomaly-detection technology then analyzes the resulting behavior to surface both expected and unexpected changes.
That approach is intended to become increasingly important as AI agents take on longer and more complicated tasks.
Raindrop cited research showing that the length of tasks AI agents can autonomously complete is doubling approximately every seven months, while individual agent runs can now last for days and involve thousands of tool calls.
Raindrop’s existing production-monitoring platform analyzes agent trajectories for semantic anomalies, allowing engineering teams to identify when behavior changes, when a problem began and which users were affected.
Examples of failures include hallucinated answers, improper tool usage, agents becoming stuck in loops and behavioral changes caused by model upgrades.
The company was founded by Zubin Koticha, Ben Hylak and Alexis Gauba. Its team includes engineers with backgrounds in fraud-transformer models at Robinhood, malicious anomaly detection at Square and security work at companies including Segment, Semgrep and Socket.dev.
Raindrop said Simulations is currently available in research preview.
KEY QUOTES:
“Agents now run for hours, call thousands of tools, and handle real money, real health data, and real customers.”
“Raindrop catches those failures in production, and with Simulations, before a change ever ships.”
Zubin Koticha, CEO of Raindrop
“Agents are fundamentally different from traditional software. They are highly capable, autonomous, and non-deterministic.”
“Raindrop treats agent failure as a detection problem, the way a security company would, and Simulations close the loop from production to development.”
Reid Christian, General Partner at CRV
