Company: Manifold
Co-Founder and CEO: Neal Swaelens, Oleks Yaremchuk
Website: https://www.manifold.security
About Manifold
Manifold secures the AI agents an enterprise workforce runs, wherever they run: on laptops, in the browser and in the cloud.
The story starts before Manifold. Our founders, Neal Swaelens (CEO) and Oleks Yaremchuk (CTO), co-founded Laiyer AI and built LLM Guard, the most widely adopted open-source LLM firewall. With more than 12 million downloads, it is still one of the most popular and most cited open-source projects in AI security. Protect AI acquired Laiyer in January 2024. There, Neal and Oleks built the first product to combine AI runtime session visibility with policy enforcement and sold it into Fortune 500 companies. Palo Alto Networks then acquired Protect AI for 750 million dollars, and inside Palo Alto they led product and engineering for a new AI agent security product.
That is where they saw the gap. The first wave of AI security tools, including the ones they had built, inspected prompts and model responses one at a time. That breaks down for an agent that logs in as an employee, runs commands and pulls in new tools on its own. In our data, prompts and responses make up 11 percent of agent events and tool calls 51 percent, so most of what an agent does never reaches a prompt filter. Security teams were stuck in the middle: boards wanted agents everywhere, and the tools on hand could only say no. Nobody was securing what agents actually do.
So in December 2025 they started again. Manifold launched publicly in March 2026, backed by a 15 million dollar seed round led by Costanoa Ventures with Cherry Ventures, Rain Capital and Modern Technical Fund. Joe Sullivan, former CSO at Cloudflare, Facebook and Uber, was among our first backers and now sits on our board.
The bet is paying off. Manifold already secures Fortune 500 companies across Europe, the US and Asia, and has won head-to-head technical evaluations against the largest security incumbents.
A lightweight collector, deployed in minutes with no proxy, gives security teams one live inventory of every agent their employees use, from coding agents like Claude Code and Cursor to browser agents and computer use. Every MCP server, skill and extension those agents load is scored by Manifest, our public AI supply chain research. Manifold ranks the risk each agent carries, follows full sessions at runtime with more than 45 threat detections, and responds in proportion, from asking the employee to confirm, to steering the agent back on track, to blocking the action outright.
Our mission is to make security the reason enterprises can adopt AI faster. At one customer, each new AI tool request took two to three hours of review and 40 to 50 days to approve. Manifold replaces that queue with policies set once and applied to every agent, so routine requests clear on their own and the security team reviews only the exceptions. Security teams go from blocking the newest AI to rolling it out first.
Want to find out what TechRound can do for your business?
