Rippling shipped the product it was sued over the moment both suits vanished

Runlayer and Rippling dropped their lawsuits against each other on Wednesday night, with no settlement, no money changing hands, and not even lawyers' fees, according to court documents seen by TechCrunch ^[1]. Rippling marked the occasion by immediately releasing its MCP gateway, the product at the heart of the dueling suits and the one that competes with Runlayer's ^[2].

The sequence matters more than the pleadings. According to Runlayer's lawsuit, Rippling tested Runlayer's MCP gateway for more than a year, with the two engineering teams working closely together, and never signed on as a customer ^[3]. Runlayer's founder Andrew Berman then received a text from a Rippling employee saying his employer was building its own MCP gateway and planned to ship it as a product, which the employee described as a clone of Runlayer's ^[4]. Runlayer sued, claiming Rippling violated contractual agreements covering the product tests ^[5]. Rippling countersued over patents, a move Runlayer read as an attempt to run up its legal bill and force a withdrawal ^[6]. Runlayer dropped its case after three weeks of discovery, and Rippling dropped its own ^[7].

Nothing was adjudicated and nothing was licensed. Because both sides walked away without a settlement or a payment ^[1], neither bought peace: the contractual theory and the patent theory both remain unresolved ^[14].

The product is not exotic, which is the point. An MCP gateway sits between an enterprise's AI agents and its software systems, retrieving data on request rather than handing agents direct access, and then layering on role-based access control, logging and usage trails ^[8]. Rippling, whose historic business is payroll and benefits management ^[9], has moved into two adjacent markets in a span of weeks: an AI gateway that routes across models and dashboards token spend by employee, putting it against Stripe, Ramp and Databricks ^[10], and now an MCP gateway that ties AI access to employee roles, putting it against Runlayer, Docker and Amazon Bedrock ^[11]. Runlayer, which came out of stealth in November 2025 and has raised $42 million from investors including Khosla Ventures' Keith Rabois and Felicis ^[12], is selling a broader bundle around the gateway, from agent creation to finding shadow AI agents running inside an enterprise without IT's knowledge ^[13].

For operators, the design-partner arithmetic is what changed. TechCrunch's read is that the long technical shoot-outs enterprises like to impose on startups need rethinking, because a buyer's needs can shift drastically between the start of an evaluation and its end ^[15]. The harsher version: a multi-quarter evaluation is a detailed specification handover, delivered for free, to an engineering organisation that can now build the thing. Berman is a third-time founder whose previous companies include Nanit and Vowel, which sold to Zapier in 2024 ^[16], and that experience did not prevent the outcome. The relevant math is that a test running more than a year, ending on or near this month, began well before Runlayer's November 2025 launch ^[17], which is to say the entire stealth period was spent teaching a prospect the product.

What to watch: whether Runlayer's differentiation holds now that the gateway itself is table stakes and the pitch rests on shadow-agent discovery and agent creation ^[13]; whether Rippling's patents reappear against anyone else, given nothing was settled ^[14]; and whether enterprise pilots in agent tooling start getting shorter, paid, or papered with real build restrictions.