The first wave of AI agents is already inside the enterprise, and the security teams are just now catching up. Runlayer, a New York-based startup, is betting that the Model Context Protocol (MCP) will be the connective tissue for this new generation of tools. Its pitch is simple: give IT a control plane to manage, secure, and observe every AI agent connection, or watch shadow IT explode. The company emerged from stealth in late 2025 with $11 million and a roster of early adopters that suggests the problem is real, and the buying window is open [TechCrunch, Nov. 2025].
The MCP Security Wedge
Runlayer’s product suite is built around a single premise: MCP, an open protocol for connecting AI agents to data sources and tools, will become a standard. The risk is that without governance, any employee can connect an agent to any internal service. Runlayer positions itself as the enterprise-ready layer on top, offering what it calls an "Okta-style" directory for MCP servers [news.aibase.com, 2026]. This allows IT to pre-approve servers and map them to employee identities, creating a one-to-one permission chain.
The platform’s technical claims focus on real-time threat detection. Its ToolGuard technology is said to inspect every MCP call, blocking credential exfiltration or prompt injection attempts in under 100 milliseconds [AI CERTs News, 2026]. For existing deployments, OpenClaw Watch scans corporate device fleets via MDM integrations to find unauthorized, or "shadow," MCP servers [ClawHosters, 2026]. The goal is to turn OpenClaw and similar agent frameworks from a shadow IT risk into a managed, audited corporate tool [techbuddies.io, 2026].
Why Unicorns Are Buying First
Traction is the strongest signal in Runlayer’s early story. Within four months of a stealth launch, the company reported signing dozens of customers, including eight unicorn or publicly traded companies [AI News, 2026]. The named list is a who’s who of modern tech operators: Gusto, Rippling, dbt Labs, Instacart, Opendoor, and Ramp [TechCrunch, Nov. 2025].
These are precisely the companies moving fastest on AI adoption, where the friction between developer velocity and security compliance is most acute. A case study from Gusto claims the payroll platform went from zero to 1,500 daily AI users in 90 days using Runlayer [llms.runlayer.com]. The value proposition isn’t just stopping attacks; it’s enabling safe adoption at speed.
The Team Behind the Bet
The founding team brings a blend of entrepreneurial scar tissue and deep platform AI experience. CEO Andrew Berman is a third-time founder, having previously founded baby-monitor company Nanit and meeting platform Vowel, which sold to Zapier in 2024 [TechCrunch, Nov. 2025]. At Zapier, he led AI efforts, an experience that directly informed Runlayer’s focus. Co-founders Tal Peretz, Vitor Balocco, and Michał Wysocki were key engineers behind Zapier’s own AI agent initiatives [LinkedIn, 2026].
Their collective experience building at scale is underscored by a critical advisory relationship: David Soria Parra, the lead creator of the MCP specification itself, is an advisor to the company [TechCrunch, Nov. 2025]. This provides technical validation and a direct line to the protocol’s evolution.
| Founder | Role | Key Background |
|---|---|---|
| Andrew Berman | CEO | Founder of Nanit & Vowel (acquired by Zapier); led AI at Zapier [TechCrunch, Nov. 2025] |
| Tal Peretz | Co-Founder | Staff AI Engineer at Zapier; alumnus of IDF Unit 8200 [TechCrunch, Oct. 2025] |
| Vitor Balocco | Co-Founder | Staff Software Engineer at Zapier, worked on Zapier Agents [Crunchbase] |
| Michał Wysocki | Co-Founder | Staff Engineer, Applied AI; built Zapier Agents with Peretz and Balocco [LinkedIn, 2026] |
The Competitive Landscape
Runlayer is not planting a flag in empty territory. The space for securing AI deployments is attracting attention from giants and startups alike. The company’s own materials position it against Webrix, an MCP gateway provider [Runlayer]. But the broader competitive set includes infrastructure and security heavyweights like Cloudflare, Docker, and Wiz, all of which could extend their existing platforms into this new layer.
The startup’s answer to this pressure rests on three points:
- First-mover credibility. Signing eight unicorn reference customers before most competitors have a product creates a proof-of-concept moat.
- Protocol-native focus. With MCP co-creator David Soria Parra as an advisor, Runlayer is betting on deep integration with the standard’s roadmap.
- Team density. The founding team’s combined experience in building AI at scale and previous founder exits is a tangible asset for enterprise buyers evaluating a new vendor.
The real competition may be less about features and more about which layer of the stack becomes the primary point of control. Will security teams want a standalone MCP security console, or will they demand these capabilities be baked into their existing cloud security posture management (CSPM) tools?
The Capital and the Clock
Runlayer’s $11 million seed round was led by Khosla Ventures, with general partner Keith Rabois taking a personal interest, and Felicis [SecurityWeek, 2025]. The round values execution speed. The capital is a bet that Runlayer can establish itself as the default enterprise choice before the market consolidates or before larger players decide to build rather than buy.
The next twelve months will be about conversion. The company has landed flagship logos; the question is whether it can translate those early deployments into seven-figure annual contracts and expand within those accounts. The seed funding provides a runway to build out sales and customer success while the product matures. For Khosla Ventures and Felicis, the bet is that securing the agent layer will be a foundational, venture-scale business. The real test for Runlayer is whether the enterprise’s appetite for agent security grows as fast as its appetite for the agents themselves.
Sources
- [TechCrunch, Nov 2025] MCP AI agent security startup Runlayer launches with 8 unicorns, $11M from Khosla’s Keith Rabois and Felicis | https://techcrunch.com/2025/11/17/mcp-ai-agent-security-startup-runlayer-launches-with-8-unicorns-11m-from-khoslas-keith-rabois-and-felicis/
- [SecurityWeek, 2025] Runlayer Emerges From Stealth Mode With $11 Million in Funding | https://www.securityweek.com/runlayer-emerges-from-stealth-mode-with-11-million-in-funding/
- [AI News, 2026] Runlayer signs dozens of customers including eight unicorns | https://www.artificialintelligence-news.com/2026/01/15/runlayer-signs-dozens-customers-eight-unicorns/
- [llms.runlayer.com] Gusto case study on AI user growth | https://llms.runlayer.com/
- [news.aibase.com, 2026] Runlayer's Okta-style directory for MCP servers | https://news.aibase.com/
- [AI CERTs News, 2026] ToolGuard blocks threats in under 100ms | https://aicerts.news/
- [ClawHosters, 2026] OpenClaw Watch scans for shadow MCP servers | https://clawhosters.com/
- [techbuddies.io, 2026] Runlayer turns OpenClaw into a managed corporate tool | https://techbuddies.io/
- [TechCrunch, Oct 2025] Profile on Tal Peretz's background | https://techcrunch.com/2025/10/27/israeli-intelligence-vets-raise-20m-to-track-developer-buying-signals/
- [LinkedIn, 2026] Post on founding team and Michał Wysocki's role | https://www.linkedin.com/
- [Runlayer] Competitive comparison with Webrix | https://llms.runlayer.com/blog/runlayer-vs-webrix