MCP startup Runlayer accuses Rippling of stealing its product idea | TechCrunch

Runlayer, a startup that raised $42 million from Khosla Ventures and Felicis to build a secure gateway for the Model Context Protocol, filed suit against HR software giant Rippling on July 28, 2026, alleging the company used nearly a year of intimate technical access to clone its core product. The lawsuit claims trade secret misappropriation and breach of contract — and it arrives with a striking piece of alleged evidence: a text message from a Rippling insider telling Runlayer's CEO that the company was building a "1:1 copy" of its gateway.

The Claim

Runlayer's product sits at a chokepoint in the emerging AI agent ecosystem. MCP — the Model Context Protocol, originally developed by Anthropic — provides a standardized way for AI models to call external tools and services. A gateway in this context is the security and routing layer between an enterprise's internal systems and the flood of MCP-compatible AI agents that want to reach them. Runlayer's pitch to enterprises was that it could manage authentication, access control, auditing, and rate limiting across all those connections from a single control plane.

According to the lawsuit, Rippling was not merely evaluating this product — it was deeply embedded in it. The company spent close to a year in trial with Runlayer, and Runlayer alleges that access during that period extended to source code, internal roadmaps, and sustained engineering collaboration between the two teams. When the two companies failed to reach agreement on pricing for a full enterprise license and Runlayer ended the trial, a person inside Rippling allegedly reached out directly to Runlayer's CEO to warn that Rippling had pivoted to building the same product in-house — nearly identically.

Rippling does not deny building an MCP gateway. The company frames it as a natural product extension for an HR platform increasingly expected to integrate with AI agents, and dismisses Runlayer's claims as manufactured grievances from a startup trying to suppress legitimate competition. In Rippling's telling, any architectural resemblance between the two products is simply the result of working in a well-defined technical space, not copying.

What We See

The text message from the Rippling insider is the most concrete piece of evidence Runlayer has surfaced publicly, and it is genuinely unusual. Internal employees at enterprise software companies do not routinely contact outgoing vendor CEOs to warn them about in-house builds — not in ways that end up attached to lawsuits. If its contents hold up under scrutiny, the communication shifts the character of this case from "two companies built similar things" to "one company knew exactly what it was building toward."

Our read is that the core tension here is not whether Rippling can build an MCP gateway — a company of its scale obviously can — but about what it absorbed during a trial that included source code and internal roadmaps. Those are not standard evaluation terms. Normal enterprise software trials involve sandboxed environments, scoped feature access, and demo data. Sharing source code during a vendor evaluation is the kind of risk experienced enterprise software sellers increasingly refuse to take, for exactly this reason: the evaluating party internalizes the architecture and builds it themselves. That Runlayer extended that access suggests either strong confidence in its legal agreements or a significant misjudgment of how much protection those agreements actually provide.

Multiple TechCrunch sources confirm the story broke July 28 and was filed under the Startups category — not AI or Enterprise. That framing is worth noting: this is as much a startup-versus-incumbent story as a technical IP dispute. Rippling has tens of thousands of enterprise customers and can offer an MCP gateway as a line item on a renewal conversation. Runlayer has to sell it as a standalone product to each new buyer. The $42 million from Khosla and Felicis would have positioned Runlayer as a category leader — if the category holds together long enough for standalone vendors to win it. The lawsuit suggests it might not.

Where It Falls Short

What the lawsuit and initial reporting leave unresolved is whether the MCP gateway space has enough technical depth to support trade secret protection at all. If Rippling can produce a comparable product after a year of exposure, that does not automatically prove theft — it might mean the core product is more straightforward to build than Runlayer's fundraising narrative implied. Trade secret cases typically require showing that the information taken was genuinely non-obvious to a skilled practitioner working independently. Rippling's defense will almost certainly argue that any competent engineering team working from the MCP specification and standard enterprise security patterns would arrive at similar architectural decisions, no source code required.

The insider text message complicates that defense, but a single informal message from an employee is not the same as documented engineering decisions traceable to Runlayer's proprietary materials. Courts have seen cases where warnings of this kind reflected internal speculation rather than executive instruction. Runlayer will need discovery to establish whether Rippling's architecture contains specific, non-obvious implementation choices — the kind of fingerprint that appears when you build from someone else's source rather than from a protocol specification. Until that evidence surfaces, both sides are in the familiar pre-discovery position where the party with the better press narrative holds the early advantage. Right now, that is Runlayer.

Sources