About
We're building the layer AI agents run on
Arclight started because our own team kept losing trust in the agents we shipped — not because the models were bad, but because we had no way to see what they were actually doing.
In 2022, our founding team was running agent workflows in production for the first time — and quickly realized the hardest part wasn’t building the agents. It was trusting them enough to leave them unattended.
A bad deploy could quietly touch data it shouldn’t have. A reasoning loop could go sideways with no clear point to intervene. The tools we had were built for deterministic software, not probabilistic agents making judgment calls in real time.
Arclight is the platform we wished we’d had — a place to define what agents are allowed to do, watch what they actually do, and step back in the moment something goes wrong.
2022
Founded
38
Team members
1,200+
Teams on Arclight
$28M
Raised to date
What we believe
Principles that shape the product
Traceability over trust me
We believe engineering teams shouldn't have to take an agent's word for it. Every decision should be inspectable.
Guardrails, not gatekeeping
Safety controls should make teams move faster with confidence, not slower with bureaucracy.
Production is the only demo that matters
We build for what breaks at 2am with real users, not what looks good in a five-minute walkthrough.
Boring on purpose
Infrastructure should be predictable. We'd rather be dependable than flashy.
Team
The people building Arclight
Priya Nair
Co-founder & CEO
Marcus Ohene
Co-founder & CTO
Elena Vasquez
Head of Engineering
Tomás Ferreira
Head of ML Infrastructure
Sofia Lindqvist
Head of Platform Reliability
Daniel Okoye
Head of Design