A Funding Round That Turned Into a Fight
On June 24, 2026, Runlayer announced a 30 million dollar Series A led by Felicis, with Khosla Ventures participating, lifting the company's total raised to 42 million dollars after an 11 million dollar seed. The detail that made the round newsworthy was not the size but the appetite. Fortune reported that Vinod Khosla wanted to buy every available dollar of the round when he heard Runlayer was raising. That kind of crowding only happens when investors believe a category is forming, and the category here is governance for the agent workforce. We read the enthusiasm less as a verdict on one startup and more as a bet that someone, eventually, has to sit between enterprise data and the agents reaching for it.
Felicis general partner Jake Storm, who led both rounds, put the thesis bluntly. "A lot of people view governance as a tax. This is actually the unlock. It flips it totally on its head." That framing matters for executives who have spent two years treating AI controls as a brake on adoption. Storm went further on why a startup, not an incumbent, could win the layer: "This is a Switzerland business. No platform can own this, a neutral, cross-provider control layer is absolutely critical if we actually believe in the future of agents performing work." Whether neutrality is defensible is the open question we return to below.
What Runlayer Actually Sells
Strip away the funding theater and Runlayer is infrastructure built around the Model Context Protocol, the open specification that defines how agents plug into enterprise systems and data. The product behaves like two things at once: a corporate app store where employees pull pre-approved AI tools with guardrails attached, and a control room where security teams see what agents are doing, what data they touch, what they cost, and where shadow AI is creeping in. The company describes its aim as a golden path for AI, meaning enablement, security and control in one platform rather than three competing tools bolted together after the fact.
The capability list reads like a CIO wish list for the agent era: identity scoping by tool, approval workflows, runtime risk checks, audit logs and real-time visibility tied to every action, all model-neutral across IDEs, chat clients and vertical apps including Salesforce Agentforce. Founder Andrew Berman, a three-time founder who previously cofounded Nanit and Vowel and ran AI at Zapier, summarized the mission in a line that should resonate with anyone accountable for an agent fleet. "You need a single pane of glass that watches everything. Who watches the watchman? That's what we do." The promise is appealing precisely because most enterprises cannot currently answer that question.
The Customer List Is the Real Signal
Pitch decks are cheap, logos are not. Runlayer says its customers already include Instacart, Gusto, Opendoor, dbt Labs, AngelList and Lemonade, alongside unnamed Fortune 500 firms including a large bank. That spread matters because it crosses regulated and unregulated sectors, which is where governance demand actually bites. A consumer app testing agents can tolerate ambiguity about data access. A bank cannot. The presence of financial-services buyers suggests the pain is acute enough that risk-averse organizations are willing to insert a young vendor into the path between agents and production systems.
We would caution executives against reading early traction as proof of durability. The same logos that validate the category today will be the first to consolidate spend tomorrow if the application platforms ship comparable controls natively. What the customer list does establish is that the problem is real and present, not theoretical. Enterprises are deploying agents faster than they can describe what those agents are permitted to do, and they are buying stopgaps to close the gap. That is the demand Runlayer is monetizing, and it is the same demand every platform vendor is now racing to claim.
The Neutrality Bet, and Its Weakness
The Switzerland argument is seductive and also fragile. The case for a neutral layer is that no single vendor should hold the keys to every agent's behavior, because agents will increasingly span Microsoft, Salesforce, ServiceNow, custom builds and open frameworks at once. A control plane owned by any one of those vendors is structurally conflicted. Runlayer's wager is that buyers will pay for an independent referee rather than trust the players to officiate their own match. For multi-vendor estates, which is most large enterprises, that logic holds up well enough to justify a line item.
The weakness is that platform incumbents rarely cede control planes quietly. Microsoft has already made governance the gate for enterprise agents, and ServiceNow is positioning its platform as the place to govern every agent regardless of where it runs. Each will argue that the safest control point is the one closest to the data and the workflow, not a separate layer customers must integrate and maintain. The honest read for CIOs is that both models will coexist for a while, and the right choice depends on how fragmented your agent estate is. The more vendors you run, the more a neutral layer earns its keep.
What CIOs Should Take From This
The lesson is not that you should buy Runlayer. It is that the market has now priced agent governance as a platform, not a feature, and that repricing should change how you budget. If venture investors are fighting over a control-layer startup, your own agent sprawl is almost certainly further along than your governance, and the gap is a liability you are carrying whether or not it appears on a risk register. Berman's framing of every employee delegating work to swarms of agents is a marketing line, but the underlying trajectory is not in dispute, and unmanaged delegation is how data leaves the building.
Our advice is to treat the control plane as an architectural decision to make deliberately now, before the agent count makes retrofitting expensive. Decide whether your strategy is platform-native governance, a neutral overlay, or a hybrid, and decide it at the enterprise level rather than letting each team improvise. The vendors competing for this layer are betting that the answer is worth tens of billions of dollars across the market. The least defensible position is the one most enterprises currently occupy, which is no answer at all and a growing fleet of agents nobody is watching.
