MCP startup Runlayer accuses Rippling of stealing its product idea

Selling software to other tech companies always carries a risk: what if the customer decides to just build it themselves after seeing how you did it? That's the exact fight now playing out between Runlayer, a startup building a secure gateway for the Model Context Protocol, and Rippling, the HR software company that reportedly spent close to a year evaluating Runlayer's product before walking away.

According to the complaint, Runlayer opened up its playbook during that trial — product roadmap, source code, the works — under a mutual NDA and a trial agreement that barred Rippling from copying its IP or building derivative products. Runlayer's founder and CEO, Andrew Berman, says the deal eventually collapsed over price. Then, per the lawsuit, a source inside Rippling reportedly texted Berman to say the company was building what amounted to a near-identical clone of Runlayer's product internally.

Runlayer is now accusing Rippling of trade secret misappropriation, unfair competition and breach of contract, and has hired Sullivan & Cromwell to press the case. Rippling isn't backing down. A spokesperson told TechCrunch the company is launching its own MCP gateway built entirely on proprietary work, and dismissed Runlayer's suit as a distraction from the startup's own business struggles. Whether Runlayer's claims hold up in court is anyone's guess, but the optics of a heavyweight law firm attached to the case do lend it some weight.

What makes this dispute worth watching isn't just the accusation itself — it's what it reveals about the messy reality of selling AI infrastructure to enterprises, especially ones stacked with engineers who could build a competing tool if the price doesn't work out. MCP itself is barely a year old, launched by Anthropic in November 2024 as an open standard for letting AI models pull in outside data and tools. Gateway products built on top of it, like Runlayer's, add the security and control layer enterprises want. Runlayer entered that market last year and has raised $42 million from backers including Khosla Ventures and Felicis, but the field has gotten crowded fast.

The uncomfortable truth here is that both sides have legitimate incentives that point in opposite directions. Startups need deep, hands-on trials to prove their product works at enterprise scale, but those same trials mean handing over the crown jewels to a potential buyer who might just as easily become a competitor.