Introducing Raindrop Triage. Today Raindrop is launching Raindrop Triage, an agent that investigates your AI agents. It lives in Slack and Web. It's also an MCP. And it's the foundation of what comes next at Raindrop. "The triage agent is great for the messy part of understanding agent behavior. You can ask quick ad hoc questions, spot broader patterns, and jump into the traces that show what's actually happening." Kaloyan Videlov AI Engineer at Framer "We have Triage running a Monday digest of our top issues and new patterns. It surfaces things we would've missed otherwise and points us to the relevant troublesome conversations." Tuan Ahn Pham AI Engineer at Speak The problem with finding agent issues. Picture a normal Tuesday. A customer tweets that your agent is broken. Someone screenshots it into Slack. An engineer opens your dashboard, starts digging through traces, and goes hunting for the pattern. Hours later, they post in the thread: "yeah, the agent's making up order IDs when the lookup tool times out." Then someone has to actually go fix it. Most of that time wasn't spent fixing anything. It was spent looking through dashboards, filtering, investigating, assembling datasets, finding the pattern, figuring out how many people are affected, and whether it's still happening. With Raindrop Triage, all of that boils down to a single question (without leaving Slack). Triage in Slack. Just @raindrop from any channel and ask it anything you'd ask a teammate who lives in your data: * @raindrop every Monday, summarize the biggest issues enterprise customers had * @raindrop how's the rollout on flag checkout-v2? * @raindrop is this still happening? (with a screenshot of a customer complaint pasted in) Triage replies in thread, keeps context across follow-ups, and you can keep poking. "Didn't we fix it last week?" "Yes, but it regressed." You can schedule automated briefs. You can run experiments. You can point it at customer complaints, Raindrop issues, or anything else you want investigated. And you never have to leave Slack. Triage Agent as an MCP. Triage isn't just a Slack bot. It's also available through Raindrop MCP. Your coding agent can use Triage as a first-class citizen. This wasn't an afterthought for Raindrop. It's laying the foundation for what Raindrop call self-healing agents. * Set up a Devin cron * "Every night, ask Raindrop for the most important issue and fix it" > Devin "check raindrop for the top enterprise issue and fix it" Calling raindrop... Found: list_projects tool fails when filtered by keyword (43 conversations affected, started April 22) | Reading src/tools/projects.ts | Editing src/tools/projects.ts, removing keyword from tool schema | Running tests... all passing | Opening PR #847 Claude Code, Cursor, Codex, Devin: they can all @raindrop now. Your coding agent pulls a live issue from production, reads the failing examples, edits the file, runs the tests, and opens the PR. Why Raindrop made its agent an MCP tool. At first glance, it might seem strange to launch its agent as an MCP tool. Why not just do what everyone else does: expose the underlying tools through the MCP and let other agents call them directly? That would've been cheaper and simpler for Raindrop (and Raindrop do also expose those tools!). But making the agent itself available through the MCP has a few special advantages. It's also setting the scene for what Raindrop call the Multi-Agent Covenant (coming soon). 1. Task-specific harness The first, and probably most obvious: Raindrop has tuned Triage's harness to be very good at debugging agent issues: understanding production data, navigating conversations, accurately calculating impact, etc. Its harness (generally) significantly outperforms Claude Code using the same tools (in speed, accuracy, etc.). When Raindrop tried to hill-climb performance for a specific coding agent (e.g. Devin, Codex, etc.), Raindrop found those efforts to be largely futile, and also out of its control. Exposing Triage as an MCP tool allows Raindrop to control the model-harness matching, and makes it instantly portable between any coding agent. 2. Personalization As you use Raindrop, it learns what you do and don't care about: * When you refine signals * When you give feedback on an issue * When an issue gets fixed * As you use Triage agent (through Web, Slack, and MCP) Exposing Triage as an MCP allows Raindrop to continually share these learnings with your coding agent. 3. Context handoff Imagine this flow: * Someone posts a bug in Slack * You tag @Raindrop to look into it * You find the issue * You tag @Devin to fix By exposing Triage as a tool, Devin can pick up the conversation right where you left off. No context lost. What's next. Triage is the foundation of the next generation of Raindrop. Most observability tools can't even tell you what broke. Raindrop closes the loop, from signal, to investigation, to fix. Now, your agent can fix itself.

Raindrop AI launches Trajectories for agent trace analysis. Raindrop AI has introduced Trajectories, a new feature aimed at helping teams review and search agent traces in a way that better matches how AI agents actually work. * The company says traditional trace views can be hard to use for agent debugging because they were built mainly for distributed systems and microservices. * Trajectories is designed to make agent activity easier to follow, especially when runs involve multiple steps, tool use, errors, and recovery paths. * Raindrop AI highlights two main capabilities: * agent focused trace visualization * natural language search across traces * With natural language search, users can describe the kind of trace they want to find, and the system turns that request into structured filters. * The search interface also suggests filters based on the tools, models, and signals already present in the trace data. * Trajectories offers two visualization modes: * Output size mode to show spans based on output token volume * Duration mode to show spans across a shared timeline for spotting slow steps * The interface also marks errors clearly so users can quickly identify failures and review how the agent responded afterward. * Raindrop AI has also added an Explain trajectory feature that creates an AI generated summary of a trace based on tool activity, timing, inputs, outputs, and errors. * Users can further narrow results with filters for: * signals * tools * models * tool sequence patterns * call and error thresholds * matching error text * The company says active filters remain in the URL, making filtered views easier to share with teammates. * To use Trajectories, users need to send trace data through Raindrop AI's TypeScript or Python tracing setup before viewing it in the dashboard. This launch reflects Raindrop AI's effort to make observability more practical for teams building and debugging AI agents.

Raindrop raises $15M funding for AI agent monitoring. * Raindrop raised $15 million in seed funding led by Lightspeed. * Demand is rising as AI agents take on longer and more complex tasks. * The company aims to scale its AI agent monitoring platform. Raindrop plans faster expansion after fresh capital. Raindrop raised $15 million in new seed funding. The company wants to scale its monitoring platform for AI agents. It also wants to meet rising demand from enterprises that rely on autonomous systems. The round was led by Lightspeed Venture Partners. It also included participation from Figma Ventures, Vercel Ventures, and YC. Several prominent founders from Replit, Speak, Framer, Cognition, and Notion joined the round. Raindrop positions itself as a simple idea. It calls itself "Sentry for AI agents." The team believes this category will become essential. Engineering teams now use AI agents across healthcare, finance, and consumer platforms. Many of these agents operate for hours without human intervention. Issues surface quietly and sometimes cause real harm. Raindrop wants to fix that. It wants to help teams detect problems early. It wants to offer visibility into complex behaviour that traditional tools cannot see. Why testing methods fall behind. AI agents now run longer processes. They use more tools. They complete more reasoning steps. They also interact with MCP servers. This expansion increases failure risk. The limits of legacy testing are clear. Evals cannot interpret long trajectories. They cannot inspect deep chains of decisions. Raindrop CEO Zubin Koticha noted that teams often have no idea where agents fail. Issues hide inside long reasoning paths or loops. Engineers cannot track these problems with latency graphs or token metrics. High-profile failures expose the stakes. ChatGPT reportedly told a user to stop taking medication. Air Canada faced a lawsuit after its chatbot promised refunds it could not honour. These incidents reveal how small issues become major outcomes when agents operate autonomously. Raindrop wants to prevent these situations. It wants to help teams identify subtle issues that cause major failures. How Raindrop's monitoring system works. Raindrop uses small, custom models. These models learn the shape of each AI product. They detect signals that other tools miss. Teams can observe agent behaviour across millions of events. They can also define custom alerts. Default signals include "User Frustration." A team may want more specific signals. They might create "Agent Stuck in a Loop." Another team might define "UI Aesthetic Complaints." Each signal becomes trackable. Sudden changes produce automatic alerts that resemble Sentry-style issue notifications. Raindrop also deploys its own AI agents in the background. These internal agents scan for failure patterns. They group related incidents. They build step-by-step explanations to show how an issue occurred. This process makes debugging easier. It also shortens the time between detection and resolution. Many customers also use Raindrop Experiments. This is an A/B testing system for AI agents. Teams test new tools or model changes. They see if fixes reduce incident rates. They confirm impact before rolling out updates. Raindrop already works with frontier AI teams. These teams process millions of events each day. They rely on Raindrop to keep failure rates low. Tolan CTO Evan Goldschmidt said the platform feels like an iOS crash report system. The difference is that it monitors AI capabilities, not mobile apps. Raindrop sees demand rising. Many companies now build autonomous systems. They need monitoring to maintain reliability. They also need tools that reveal hidden behaviour, not just surface-level metrics. Lightspeed Partner Bucky Moore said teams struggle with silent failures. Traditional evals provide little insight. He believes Raindrop defined the new standard for monitoring. He also said the firm is excited to support the company as it grows into the default observability layer for AI agents. This strong investor interest signals growing trust in agent-based systems. It also shows a shift. Companies now treat monitorability as core infrastructure. Founders who experienced the problem first-hand. Raindrop grew from a problem the founders faced. Zubin Koticha and Alexis Gauba built a coding agent before starting the company. They faced silent agent failures often. They tried to diagnose them. They struggled. They discovered the same issue at YC. Every team building an agent suffered the same experience. They teamed up with Ben Hylak. Hylak worked at Apple and later joined Apple's Human Interface Design team. His experience helped shape product quality and user experience. Together they built Raindrop to solve a real gap. The team wants to define observability for AI agents. They want to offer reliable systems to support complex autonomous products. They also want to help enterprises avoid the hidden dangers of silent agent errors. Why this funding round matters. The $15 million seed round gives Raindrop the resources to scale. It also offers validation from top investors. Lightspeed has backed companies like Anthropic, Snap, and Wiz. Figma and Vercel bring design and infrastructure expertise. Angels from leading AI companies add credibility and insight. This combination suggests Raindrop sits at the centre of a new shift. Monitoring is now a must-have layer for any agent-driven product. Demand will increase as companies deploy agents in production. Raindrop now plans to expand its monitoring models. It wants to support more use cases. It wants to meet enterprise demand more quickly. It also wants to strengthen its position as the leader in agent observability. To stay updated on crypto venture capital funding and market trends, visit its venture capital news section for more insight. Clinton Nwachukwu is a crypto and finance writer with an MBA in Artificial Intelligence and 6+ years of experience creating content for leading global brands. He turns complex topics into clear, actionable insights for readers worldwide. VentureBurn is a media platform covering the latest in cryptocurrency, artificial intelligence, venture capital, and the startup ecosystem. Opinions expressed on VentureBurn are for informational purposes only and do not constitute investment advice. Before making any high-risk investments in digital assets or emerging technologies, readers should conduct their own due diligence. All transactions and financial decisions are made at your own risk, and any losses incurred are solely your responsibility. VentureBurn does not endorse or recommend the buying or selling of any digital assets and is not a licensed investment advisor. Please note that VentureBurn may participate in affiliate marketing programs.