Runlayer raised a $30 million Series A to build the control layer for AI agents at work, and Vinod Khosla's push to take the whole round says more than the round size does.
When Vinod Khosla heard Runlayer was raising again, his response was blunt: he wanted every available dollar. That's not normal venture manners. It is what happens when an investor thinks the category is forming before everyone else has agreed what to call it. Khosla Ventures and Felicis ultimately led the $30 million Series A, bringing Runlayer's total funding to $42 million just seven months after the company came out of stealth with an $11 million seed round from the same two firms.
The company is not building another model or another agent framework. It is trying to become the layer underneath them: the place where a company can see which AI agents are running, what data they've touched, what they cost, and whether they did the job they were supposed to do. As Fortune reported, Runlayer works like a corporate app store and control room in one. An employee can connect OpenAI's ChatGPT, Anthropic's Claude, Salesforce's Agentforce, or a custom internal agent through a pre-approved channel where company data and guardrails are already in place. IT and security teams get one dashboard instead of a dozen loose agent deployments scattered across the business.
If you run a company, this is the part of AI that sounds dull until it breaks. A sales agent with the wrong permissions can expose customer data. A support agent can take an action nobody approved. A finance agent can rack up model costs without anyone owning the bill. The early agent boom has been sold as productivity. Runlayer is selling the less glamorous half: control.
Andrew Berman has seen this problem from close range. He previously cofounded Nanit, the AI baby monitor company, and Vowel, an AI video conferencing startup that sold to Zapier in 2024. His co-founders are Tal Peretz, who led machine learning for the Israeli Air Force, and Vitor Balocco, a former staff AI engineer at Zapier. Berman also served as Zapier's director of AI, which matters here. Zapier sits close to the messy middle of business software, where tools connect to other tools and small automations quietly become operational infrastructure.
Here's the problem Runlayer is selling against. Gartner projects that 40% of enterprise applications will include AI agents by the end of 2026, up from less than 5% in 2025. Agentic AI spending is on pace to hit $201.9 billion this year. That is a lot of software acting on behalf of companies, reaching into sensitive data, taking actions in production systems, answering customers, and creating logs that many firms still can't read in one place.
The large software platforms have noticed. ServiceNow, Workday, Microsoft, Salesforce, Google Cloud, you name it, all want agents to live inside their own products. But governance inside one suite is not the same thing as governance across the business. A ServiceNow feature can govern ServiceNow agents. It won't automatically tell you what a Claude agent, a ChatGPT workflow, and a custom internal agent did across three different systems at 2:14 p.m. on a Tuesday.
That is Runlayer's opening. It has to be neutral, or it has no reason to exist. The company needs to sit above every model, every agent vendor, and every internal tool without forcing the enterprise to pick a single AI stack. Frankly, that is the only version of this product that makes sense. Enterprises are not going to standardize their entire AI agent estate around one vendor just because governance is easier on a slide deck.
Customers are already testing that thesis. Within four months of coming out of stealth, Runlayer signed Instacart, Gusto, dbt Labs, Opendoor, Homebase, AngelList and others, according to the company's reported customer list. That mix matters more than a tidy logo row. It includes marketplace operations, payroll, developer tooling, real estate and small business software, which means the agent governance problem is not trapped in one vertical. It shows up wherever software is being asked to do more than answer questions.
Runlayer is also a founding member of the Anthropic-led AI Agent Interoperability Forum alongside OpenAI and Google. That puts it near the standards conversation while those standards are still being argued over. You don't have to overread that. A forum seat doesn't make a company the winner. But it does mean Runlayer is close to the plumbing debate, not waiting outside for the big platforms to publish the rules.
The market forecasts are huge, and you should treat them carefully. Research cited by Claw & Talon estimates the agentic AI security market at $55 billion in 2026 and projects it could reach $888 billion by 2035. Forecasts that far out in a young category are usually part analysis, part sales pitch. Still, the direction is not hard to believe. Once agents start making decisions and taking actions inside companies, someone has to answer the ugly question: who is accountable when one of them does something it shouldn't?
Khosla's urgency fits the old picks-and-shovels logic of AI investing. Don't only chase the application. Back the infrastructure that every application has to touch. Runlayer may become that layer, or the hyperscalers may fold enough governance into their own platforms to squeeze the independent players. That is the real question this round can't answer yet.
But the signal is clear enough. A $30 million Series A, seven months after stealth, with Khosla trying to take the whole thing, is not just another AI funding item. It is a bet that enterprise AI agents are moving faster than the systems built to govern them. If that gap stays open, Runlayer has a real shot. If it closes inside Microsoft, Google, Salesforce, and ServiceNow, the company will have to prove that neutrality is worth paying for.
Also read: Seltz raises $12.5 million to build the search layer that agentic AI actually needs • Meta and Microsoft just pre-bought the AI future, and the landlords are the ones with real leverage • The British Army just proved AI can compress 72 hours of war planning into one, and the race to replicate it has begun
