Manifold Raises $8M in Seed Funding to Secure Autonomous AI Agents

There is a moment in every cycle where the tools stop asking for permission and start taking action. That moment just got a security layer. Manifold Security pulled in $8M in seed funding and decided that watching AI isn’t enough anymore, you better understand what it’s doing when nobody’s looking.

Credit where it’s due. Neal Swaelens, Co-Founder and CEO, alongside Co-Founders Oleksandr Yaremchuk and Michael McKenna, didn’t just show up with a pitch deck and a dream. They walked in with scar tissue from building LLM Guard, one of the most widely adopted open source LLM firewalls out there. They’ve seen where the bodies are buried in AI security, and now they’re mapping the terrain where agents actually operate, on the endpoint, in motion, making decisions that touch code, production systems, and CI pipelines like it’s just another Tuesday.

Costanoa Ventures led the round, with Cherry Ventures, Rain Capital, and Modern Technical Fund riding shotgun. Then you’ve got Joe Sullivan and Vijay Bolina stepping in as angel investors, which is the security equivalent of having a couple of heavyweight trainers in your corner before the bell even rings. That kind of backing doesn’t just fund a company, it signals that the people who’ve been closest to the fire see what’s coming next.

Here’s the real play. Manifold isn’t chasing prompts, it’s tracking behavior. Their AI Detection and Response platform lives where these autonomous agents actually execute, giving teams visibility into which agents are running, what tools they touch, what systems they access, and what actions they take. No extra gateways, no architectural gymnastics, just an agentless approach that plugs into the infrastructure enterprises already have. Translation, less friction, more truth.

The takeaway for anyone building or betting on AI right now is simple. If your product can act, it can misact. And if it can misact at machine speed, you don’t get the luxury of reacting late. Manifold Security is leaning into that tension, building for a world where software doesn’t wait for instructions, it interprets them and moves. That shift isn’t theoretical anymore, it’s operational.