Asymptote Labs, a developer of runtime infrastructure for AI agents, has secured $5.5m in an oversubscribed seed round, betting that companies deploying autonomous agents at scale will soon need a single system of record to monitor and control them.
Tower Research Ventures, the venture arm of trading firm Tower Research Capital, led the investment. Lytical Ventures, 1984 Ventures and South Park Commons also took part. The startup plans to direct the capital towards engineering, product development, recruitment and its commercial push.
The raise comes as enterprises roll out growing numbers of AI agents across staff workstations, cloud environments and business processes.
These agents can reach sensitive data, alter code and trigger actions within critical systems, yet security and engineering teams often have little insight into how they behave or how to contain the risks they introduce.
Central to the company’s offering is Beacon, an open-source telemetry layer released under the MIT licence. It records agent activity across Claude Code, Codex, Cursor and over 20 further agent harnesses, converting that activity into a standard runtime data model built on top of the OpenTelemetry standard.
Organisations gain a consolidated view of every agent while retaining ownership of the underlying data. Developers also benefit, as insights from earlier coding sessions persist instead of being lost when a session closes.
Beacon’s adoption has accelerated quickly. Its GitHub stars rose by more than 460% during September, and the project climbed to third place on GitHub’s global trending repositories. Security teams are already running Beacon and the firm’s enterprise platform in live environments, with regulated enterprises among the earliest and most active users.
Asymptote Labs builds a shared runtime data layer that provides enterprises with real-time visibility, intelligence and control over AI agents across different tools and environments.
Its enterprise platform, which sits on Beacon, inspects agent activity as it happens and applies policy-based guardrails at the prompt and tool levels, redirecting agents away from unsafe conduct and stopping hazardous actions before they execute. An expanding set of detections flags issues within minutes of installation, spanning data leakage, insider threats, AI supply chain risks and inefficient workflows.
Justin D’Souza and Shukan Shah, both early engineers at social engineering defence firm Doppel, founded the business. At Doppel, D’Souza scaled the real-time machine learning detection platform from handling 1 million to over 1 billion domains daily, while Shah led SOC workflow automation and developed agentic threat-hunting infrastructure. The team also draws on experience from security research firm Trail of Bits.
Following Doppel, the pair joined the South Park Commons Founder Fellowship. Angel investors include Doppel co-founders Kevin Tian and Rahul Madduluri, along with Socket founder Feross Aboukhadijeh.
Although security is the initial focus, the company sees wider applications. The same telemetry layer could help engineering teams strengthen reliability, platform teams oversee adoption and FinOps teams control AI costs.
Over time, accumulated agent data could also support more capable, self-improving agents. Tower Research Ventures came across the startup through Beacon’s open-source momentum and is now testing the technology inside its own security organisation.
Asymptote Labs co-founder and CEO Justin D’Souza said, “AI agents are becoming a new operating layer inside every company, but the infrastructure to understand and control them hasn’t caught up.
“We saw this problem firsthand when building agents at Doppel. As companies deploy thousands of agents across their infrastructure, their runtime data will become one of the most valuable assets they own. We built Beacon as open source because we believe every organization should own that data, and Asymptote enables them to act on it.”
Copyright © 2026 RegTech Analyst
Copyright © 2026 RegTech Analyst
