Asymptote Labs Raises $5.5M to Show Security Teams What Their AI Agents Are Doing

retained text

Enterprises have moved from asking AI to suggest code to letting AI agents write it, query sensitive data, and call internal tools on their own. Security and engineering teams built their monitoring for human users and static software, so many of them now struggle to answer a basic question: what did that agent just do? Asymptote Labs builds the runtime infrastructure that answers it. The company’s open-source project, Beacon, captures agent activity across Claude Code, Codex, Cursor, and more than 20 other agent harnesses, then normalizes it into a common data model that extends OpenTelemetry. Its enterprise platform builds on that data to run real-time detections, enforce policy at the prompt and tool levels, and block risky actions before they happen. Beacon’s GitHub stars grew more than 460% in September and the project reached No. 3 on GitHub’s global trending repositories, while financial services firms and other regulated enterprises lead early adoption.
AlleyWatch sat down with Asymptote Labs cofounder and CEO Justin D’Souza to learn more about the business, its future plans, recent funding round, and much, much more…
Who were your investors and how much did you raise?
We raised an oversubscribed $5.5M seed round led by Tower Research Ventures, the venture arm of Tower Research Capital, with participation from Lytical Ventures, 1984 Ventures, and South Park Commons. We’re using the funding to expand our engineering team, accelerate product development, and bring Asymptote to more enterprises.
Tell us about the product or service that Asymptote Labs offers.
Asymptote builds runtime infrastructure for AI agents. Our open-source product, Beacon, captures what agents are doing across tools like Claude Code, Codex, Cursor, and more than 20 other agent harnesses, then normalizes that activity into a common data model.
Our enterprise platform builds on Beacon with real-time security detections, policy
enforcement, and governance. It gives organizations a unified view of their AI agents and the ability to detect threats, investigate behavior, and block risky actions before they occur.
What inspired the start of Asymptote Labs?
Shukan and I have spent more than seven years building together, most recently as early engineers at Doppel, where we developed large-scale security detection systems and agent-powered security automation.
We saw firsthand how powerful AI agents could be, but also how difficult they were to understand and control once they started doing real work. Agents were accessing sensitive data, invoking tools, and taking actions across critical systems, yet there was no shared infrastructure to understand what they were doing.
We became convinced that as companies moved toward deploying thousands or even millions of agents, they would need an entirely new layer of infrastructure to make that activity visible, secure, and useful. That’s what led us to start Asymptote.
How is Asymptote Labs different?
Most existing security and observability tools weren’t built for autonomous AI agents. Agent activity is fragmented across different tools and environments, making it difficult to get a complete picture of what’s happening.
We’re approaching the problem from the infrastructure layer up. Beacon is open source, works across agent harnesses, and gives organizations ownership of their runtime data rather than locking it into a proprietary platform.
That same data layer powers our enterprise security capabilities, allowing teams to move beyond simply monitoring agents to detecting threats and enforcing controls in real time. And because the data is useful beyond security, it can also help developers improve agent performance and carry learnings forward across coding sessions.
What market does Asymptote Labs target and how big is it?
We’re initially focused on enterprise AI security, particularly organizations deploying AI agents across engineering teams, employee workstations, and cloud environments. Financial services and other highly regulated industries have been among our earliest adopters because they need strong visibility and controls over systems accessing sensitive data.
But we believe the opportunity is much larger than security alone. We’re moving toward a world where enterprises may operate millions of AI agents, and every one of those agents will generate runtime data that needs to be captured, understood, and acted upon.
That creates an opportunity spanning security, observability, governance, engineering productivity, and AI cost management. We see agent runtime infrastructure becoming a foundational part of the enterprise technology stack.
What’s your business model?
We follow an open-source and enterprise software model. Beacon is free and MIT-licensed, allowing developers and organizations to capture agent activity and maintain ownership of their data.
We also offer Beacon Cloud, which makes it easy to centralize, search, and analyze agent activity without managing the underlying infrastructure.
For enterprises, we offer a commercial platform with real-time security detection, policy enforcement, governance, and advanced deployment capabilities. Our business model is built around helping organizations securely operate AI agents at scale.
How are you preparing for a potential economic slowdown?
We’re focused on building something customers consider essential rather than optional. As companies adopt AI agents, they need to understand what those agents are doing, particularly when they’re accessing sensitive information or taking actions in production environments. That need doesn’t disappear during a downturn.
We’re also keeping the team lean and being deliberate about where we invest. Our open-source approach lets developers adopt Beacon without a significant upfront commitment, while giving us direct feedback on what customers actually need. We believe staying close to users and maintaining capital efficiency gives us the flexibility to navigate different market conditions.
What was the funding process like?
We were fortunate to have strong interest from investors who understood where enterprise AI was heading. A big part of that came from the traction we were seeing with Beacon, which we launched as an open-source project in May.
In September alone, our GitHub stars grew by more than 460%, and Beacon reached No. 3 among GitHub’s global trending repositories. That momentum helped validate that developers were experiencing the same problem we were trying to solve.
Tower Research Ventures actually discovered us through Beacon, and their security team began evaluating the technology internally. Having investors engage directly with the product made the conversations especially meaningful. We ultimately closed an oversubscribed $5.5 million seed round.
What are the biggest challenges that you faced while raising capital?
One of the biggest challenges was communicating the scale of an opportunity that’s still emerging. There’s an enormous amount of competition in AI, particularly around building better models and agents, but we see that as validation of just how large this market will become.
As more companies build and deploy agents, the need for shared infrastructure to understand, secure, and improve them becomes increasingly important. We believe enterprises will use many different agents across tools and environments, and no single model provider or agent harness will own that entire ecosystem.
The challenge was helping investors look beyond individual AI products and recognize the opportunity to build foundational infrastructure that every agent will eventually need. Our open-source traction and early enterprise adoption helped make that vision tangible.
What factors about your business led your investors to write the check?
I think it came down to three things: our team’s experience, the scale of the opportunity, and the early traction we had demonstrated.
Shukan and I spent years building security infrastructure at Doppel. I scaled the company’s machine learning detection platform from processing one million to more than a billion domains per day, while Shukan led SOC automation and agent-powered threat hunting. We’d already spent years solving difficult problems at the intersection of AI, security, and infrastructure.
With Beacon, we demonstrated that we could build technology developers genuinely wanted to use. The rapid growth of the open-source project, combined with early adoption by enterprise security teams, gave investors confidence that we were addressing a real and growing problem.
With Beacon, we demonstrated that we could build technology developers genuinely wanted to use. The rapid growth of the open-source project, combined with early adoption by enterprise security teams, gave investors confidence that we were addressing a real and growing problem.
What are the milestones you plan to achieve in the next six months?
Our biggest priority is turning Beacon’s early developer momentum into a widely adopted infrastructure layer for AI agents.
We’re expanding our engineering and security research teams, growing our library of real-time security detections, and making it easier for enterprises to deploy Beacon across their environments.
We’re also investing in capabilities that help developers get more value from their agent activity, including turning previous coding sessions into reusable knowledge that improves future agent workflows.
Ultimately, we want Beacon to become the default way organizations capture and understand agent activity, with Asymptote providing the intelligence and controls built on top.
What advice can you offer companies in New York that do not have a fresh injection of capital in the bank?
Focus on building something people (or AI agents!) genuinely want before worrying too much about fundraising. The strongest signal you can show investors is that users are getting real value from what you’ve built.
For us, open source was an incredibly effective way to do that. We could put Beacon directly into developers’ hands, learn from how they used it, and build a community around the problem we were solving.
New York also has a remarkable concentration of potential customers, especially in financial services, security, and enterprise technology. Take advantage of that proximity. Get in front of users, listen carefully, and iterate quickly. Strong customer relationships and evidence of demand are valuable regardless of the fundraising environment.

document fetched 2026-10-24 11:27 · canonical https://alleywatch.com/2026/10/asymptote-labs-ai-agent-governance-runtime-infrastructure-security-beacon-open-source-telemetry-justin-d-souza