Key Takeaways

  • Runlayer alleges Rippling cloned its MCP gateway after a year-long trial that handed over source code and roadmap
  • Enterprise AI infrastructure sales require deep technical trials that effectively give competitors the blueprint
  • Rippling calls the suit panic over competition, not IP theft — the standard defense when the replicator has engineering muscle
  • MCP gateways are commoditizing fast; the real moat isn't the protocol but the enterprise trust layer

Runlayer gave Rippling the keys. Not metaphorically. Source code. Product roadmap. Nearly a year of engineering collaboration under a mutual NDA and a trial agreement that explicitly forbade derivative works. Then the price talks collapsed. Runlayer walked. Weeks later, a text from a Rippling insider: they're building a clone. One-to-one.

This is the hazard of selling AI infrastructure to companies that can build it themselves.

The Model Context Protocol is open source. Anthropic published it in November 2024. It solves a real problem: letting models and agents securely reach outside data and tools. But the protocol is the easy part. The hard part is the gateway — the control plane, the audit trails, the policy engine, the enterprise glue that makes security teams sign off. That is what Runlayer built. That is what Rippling allegedly copied.

Rippling denies it. Their statement is a masterclass in the genre: "panic," "fabricating claims," "business failures," "superior product," "proprietary information." Notice what's missing. No denial that they saw the code. No denial that the trial ran nearly a year. No denial that the products look alike. They argue independent creation. That argument gets harder when the independent creation arrives weeks after the trial ends.

Sullivan & Cromwell represents Runlayer. White-shoe firms don't file frivolous trade-secret suits. They file suits they think survive a motion to dismiss. That doesn't guarantee a win. It guarantees the complaint isn't theater.

The broader story is the trial trap. Enterprise sales cycles for complex Infrastructure run long because buyers demand proof. They want hands-on access. They want to stress-test the architecture. They want the engineering team in the room. For a startup, that access is the sale. For a buyer with engineering muscle, that access is a free architecture review. The NDA and the trial agreement are paper. The code is already in their repo.

MCP gateways are crowding. Anthropic, Microsoft, a half-dozen startups. The protocol is commoditizing in real time. The differentiator was never the spec. It was the enterprise hardening — the role-based access, the data-loss prevention, the compliance artifacts that let a CISO sleep. Runlayer invested there. Rippling allegedly harvested there.

The lawsuit will hinge on forensic code comparison. Variable names. Error messages. Architectural quirks that don't emerge from independent implementation of a public spec. If Runlayer's engineers left fingerprints, Rippling's "proprietary information" defense evaporates. If the products merely converge on the same optimal design, Runlayer loses. Trade-secret law doesn't protect obvious solutions.

But the cautionary tale stands regardless. Startup sells AI plumbing to tech company. Tech company has 200 engineers. Startup hands over crown jewels to close the deal. Deal dies. Crown jewels reappear in competitor's catalog. The NDA gets litigated. The startup burns runway on lawyers. The market moves on.

Runlayer's mistake wasn't the trial. It was the absence of a contractual kill shot — a clause that converts the trial into a paid license or a binding term sheet if the buyer builds a competing product within a defined window. Most startups don't have the leverage to demand it. Rippling likely wouldn't have signed it. That tension is the point.

The MCP layer will consolidate. The winners won't have the cleverest protocol implementation. They'll have the SOC 2 reports, the FedRAMP authorization, the customer references that prove the gateway survives a penetration test. Runlayer may have had those. Rippling is betting they can build them faster than Runlayer can litigate.

That bet works until it doesn't. Enterprise buyers notice lawsuits. They notice when a vendor's core IP is contested. They notice when the "superior product" ships with a cloud of deposition transcripts. Rippling's HR software customers are watching. So are the VCs who funded both sides.

The protocol is open. The trust is earned. Runlayer tried to sell trust. Rippling allegedly tried to shortcut it. The court will decide the law. The market has already decided the lesson.