Runlayer and Rippling’s MCP Gateway Fight Ends Without a Settlement

The Runlayer–Rippling dispute ended almost as quickly as it began. On August 20, 2026, the companies jointly dismissed their competing lawsuits with prejudice. Court filings said there would be no compensation, no reimbursement of legal costs, and no settlement between them. Rippling then released the MCP gateway at the center of the conflict. 101

That ending resolved the litigation, not the accusations. No court ruled on whether Rippling misused Runlayer’s confidential information or whether Runlayer infringed Rippling’s patents. The commercial contest moved forward instead.

What happened in the dispute

Runlayer sued Rippling on July 28, alleging that the larger software company had evaluated its MCP gateway for more than a year, worked closely with its engineering team, and then used confidential information to develop a competing product rather than becoming a customer. Rippling denied the allegations. 1417

Rippling responded with a patent lawsuit alleging that Runlayer infringed three Rippling patents. It also said it was developing its own MCP gateway, which would compete directly with Runlayer’s product. 7

The cases lasted roughly three weeks before the parties filed joint stipulations of dismissal. Because the dismissals were with prejudice, the specific cases were ended permanently, but the filings did not establish that either side’s allegations were true. Each company was left to bear its own legal costs. 10

Why Rippling’s launch mattered

Rippling launched its competing gateway on the same day the lawsuits disappeared. That timing made the practical outcome clearer than the legal one: the product Runlayer had sought to block entered the market, while Runlayer remained a direct competitor. 114

Rippling positioned the gateway as part of a broader AI-governance offering within its business-software platform. Runlayer, by contrast, describes itself as a broader enterprise control plane for MCP servers, Skills, plug-ins, AI clients, and agents. Its published product materials include an MCP gateway, policy and OAuth controls, runtime security, audit trails, an agent runtime, and tools for finding unmanaged or “shadow” MCP deployments. 293127

The available reporting does not establish that the two products have identical architectures or feature sets. The defensible comparison is narrower: both companies are competing for the enterprise layer that governs how AI agents access tools and business data, while Runlayer presents a more specialized, multi-product control-plane model. 71429

What an MCP gateway does

The Model Context Protocol, or MCP, provides a standardized way for AI clients and agents to connect with external tools and data. An MCP gateway sits between those clients and enterprise systems, creating a centralized point where organizations can manage access and monitor activity. 2132

In practice, a gateway can help an organization:

• Authenticate and authorize requests: connect requests to user or service identity and apply role-based or least-privilege policies.
• Control tool and data access: allow approved MCP servers and tools while restricting what an agent can reach.
• Apply runtime security: inspect or filter activity as requests move between agents and tools.
• Create observability: record requests, tool calls, policy decisions, failures, and traces for investigation and audit.
• Support governed operations: keep identity, policy, security checks, and auditing attached to MCP requests before company systems are reached. 293234

Enterprise control planes may also add model routing and token-spend tracking. Those capabilities can help teams direct workloads to approved models and understand usage by employee, team, agent, or workflow. However, the sources available for this dispute do not establish that Rippling and Runlayer currently implement every one of these capabilities in the same way.

Runlayer’s background

Runlayer is a New York enterprise-AI infrastructure startup that emerged from stealth in November 2025. It launched with $11 million in seed funding from Felicis and Khosla Ventures. 47

In June 2026, the company announced a $30 million Series A led by Felicis, with participation from Khosla Ventures, bringing its reported total funding to $42 million. Andrew Berman, who previously founded Nanit and worked in AI at Zapier, leads the company as co-founder and CEO. 404547

That financing and early customer traction positioned Runlayer as a heavily backed specialist in a market that was still taking shape. The lawsuit showed the vulnerability that can come with that position: a startup may need to expose its product deeply to win a major enterprise account, even as the prospective customer gains enough knowledge to become a competitor.

The founder lesson: an evaluation can become a competitive engagement

The allegations remain disputed, but the underlying business risk is clear. A long proof of concept can involve access to product roadmaps, technical documentation, architecture, engineering workflows, and sometimes source code. If the prospect ultimately decides to build internally, the startup may have spent months training a potential rival while delaying other sales.

That risk is particularly acute in AI infrastructure. Protocols, model capabilities, security expectations, and product categories can change rapidly, reducing the value of a traditional year-long evaluation. The Runlayer–Rippling timeline compressed the dispute into weeks: a lengthy evaluation was followed by litigation, limited discovery, dismissal, and a competing launch. 1014

For startups, practical safeguards include:

1. Limit the pilot’s scope. Demonstrate only the capabilities needed to answer the purchasing question.
2. Use staged disclosure. Keep sensitive architecture, roadmap details, and implementation information behind explicit milestones.
3. Set an evaluation deadline. A pilot should have a defined end date and a documented decision process.
4. Define conversion criteria. Agree in advance on what success, procurement, and deployment would require.
5. Control technical access. Use isolated environments, granular permissions, logging, and revocable credentials.
6. Create a build-versus-buy checkpoint. Ask early whether the prospect is evaluating a vendor or gathering requirements for an internal product.

An NDA can protect specific information, but it cannot by itself guarantee that a well-funded company will not pursue the same market. Legal action may address particular conduct, yet it may not stop a competitor from shipping while the market continues to move.

What the dispute ultimately shows

The case was not a legal verdict on MCP gateway technology. It was a vivid example of how quickly an enterprise sales process can turn into a product and intellectual-property conflict.

Runlayer entered the fight as a funded specialist building a governed control plane for enterprise AI agents. Rippling emerged from it with a competing gateway launched immediately after the dismissals. With no settlement, payment, licensing agreement, or judicial decision on the merits, the decisive question is now commercial: which approach will enterprises trust to control the fast-growing connection between AI agents, tools, and company data?