Raindrop, a San Francisco–based monitoring platform designed to detect hidden failures inside AI agents, has raised $15 million in seed funding led by Lightspeed Venture Partners. The round includes participation from Figma Ventures, Vercel Ventures, the founders of Replit, Cognition, Framer, Speak, Notion, and YC. The capital will help Raindrop meet surging enterprise demand and strengthen its position as the default monitoring layer for agentic AI systems.
Raindrop is emerging as a foundational tool for engineering teams deploying increasingly complex agents that reason over long trajectories, use multiple tools, and autonomously run for extended periods. As real-world failures make headlines, including medication-related chatbot errors and high-profile customer-service agent mistakes, enterprises are seeking more reliable, production-grade observability.
Traditional monitoring tools only exposed latency, token usage, or generic toxicity metrics, leaving teams blind to deeper behavioral failures. Raindrop, instead, introduced a new standard built around small, adaptive models customized to each AI product. This approach surfaces previously invisible signals, from “User Frustration” to product-specific indicators like “UI Aesthetic Complaints” or “Agent Stuck In A Loop,” enabling teams to track incident rates across millions of events. When issues emerge, Raindrop sends Sentry-style alerts.
The platform’s background AI agents also autonomously triage and investigate patterns, collecting evidence and generating step-by-step explanations of newly discovered problems — such as instances where an agent encourages harmful behavior.
Frontier AI companies processing millions of events per day rely on Raindrop for rapid detection, monitoring, and resolution of critical issues. Many also use Raindrop Experiments, the first A/B testing framework for AI agents, to validate fixes across toolchains, models, or pipeline configurations.
Ben Hylak, Zubin Koticha, and Alexis Gauba founded Raindrop. Koticha and Gauba are second-time founders whose previous company was acquired by Coinbase, while Hylak previously served as an engineer on Apple’s Human Interface Design team.
KEY QUOTES:
“AI agents are more capable than ever. They’re reasoning longer, using more tools, and connecting to MCP servers. Some agents now run autonomously for hours. Traditional testing methods, like evals, aren’t capable of handling the complexity of these long trajectories. This leaves engineering teams in the dark — unable to discover or track issues. Raindrop is the first monitoring platform to solve this problem.”
Zubin Koticha, CEO, Raindrop
“Our founding team was building a coding agent ourselves and kept running into silent agent failures in production. When we did YC we saw everybody building agents in our batch had the same problem.”
Zubin Koticha, CEO, Raindrop (founding origin)
“We keep seeing AI engineering teams struggling with agent failures in production, and traditional evals are not really helping. We invested in Raindrop because they defined monitoring for AI agents. They made an early bet that monitoring would be the most critical part of building reliable agents, and they’ve been right. They’re growing fast as Raindrop is being embraced by some of the most forward-thinking engineering teams in AI.”
Bucky Moore, Partner, Lightspeed Venture Partners
“Raindrop has been invaluable as we’ve been growing quickly. It’s critical for us to keep issue incidence below an acceptable threshold and become aware of any spikes. It’s like if we see an iOS crash report in Sentry, but for our AI capabilities.”
Evan Goldschmidt, CTO, Tolan
