Runlayer, a pioneering **MCP AI agent security startup**, has officially exited stealth mode, announcing an $11 million seed round led by Keith Rabois of Khosla Ventures and Felicis. In just four months, the company has rapidly gained traction, securing contracts with eight unicorns, including Gusto, Instacart, Rippling, dbt Labs, Opendoor, and Ramp[1][2]. This launch not only marks a significant milestone for AI and cybersecurity, but also signals a critical shift in how enterprises approach the risks of deploying autonomous AI agents at scale.
## The Rise of MCP and the AI Agent Security Challenge
The **Model Context Protocol (MCP)** is an emerging open-source standard that enables AI agents to interact with enterprise data and systems autonomously. Developed originally by Anthropic, MCP empowers AI agents to securely perform complex workflows and access sensitive data without constant human oversight[2]. However, as adoption accelerates, MCP’s very openness has exposed enterprises to new categories of risk — including prompt injection, privilege escalation, and unauthorized data access.
Recent, high-profile security incidents involving MCP implementations at major platforms like GitHub and Asana have underscored these challenges[1][2]. For example:
| Company | Security Incident | Date |
|———-|————————————————————————–|————|
| GitHub | Prompt injection vulnerability exposing private repositories | May 2025 |
| Asana | Server-side MCP vulnerability risking customer data exposure | June 2025 |
As more organizations integrate AI agents into their operations, the need for robust, enterprise-grade security solutions for MCP has become urgent.
## Runlayer’s Comprehensive Security Approach
**Runlayer** distinguishes itself by delivering an all-in-one security platform for MCP deployments, addressing the protocol’s lack of built-in controls. Unlike many competitors that focus narrowly on perimeter defense or basic policy enforcement, Runlayer offers a suite of features purpose-built for the realities of autonomous agent operations:
– **Advanced Threat Detection:** Every MCP request is analyzed for malicious or anomalous behavior, leveraging AI-driven detection techniques.
– **Comprehensive Observability:** The platform provides detailed, real-time monitoring across all MCP-enabled servers, giving security teams visibility into agent actions and data flows.
– **Granular Permissions Management:** Integrates with identity providers such as Okta and Microsoft Entra, enabling fine-grained access policies that align with enterprise security requirements.
– **Enterprise Development Tools:** Supports the creation, testing, and deployment of custom AI automations in a secure, controlled environment[1][2].
This holistic approach ensures that enterprises can both harness the power of AI agents and protect against the emerging attack surface they introduce.
## Leadership, Vision, and Strategic Partnerships
Runlayer’s founding team brings deep credibility and experience to the MCP security space. CEO **Andrew Berman** is a serial entrepreneur with a track record of building and exiting successful tech startups, including Nanit and Vowel (acquired by Zapier)[2]. Notably, Berman previously served as Zapier’s AI director, where he built one of the first MCP servers deployed in the enterprise[1].
Perhaps most significantly, **David Soria Parra** — the original creator of MCP at Anthropic — has joined Runlayer as an angel investor and advisor. This direct connection to the protocol’s origins gives Runlayer unique insight into its design, evolution, and potential vulnerabilities, allowing for more proactive and effective security solutions[1][2].
## Investor Confidence: Backing from Keith Rabois and Felicis
The $11 million seed round, led by **Keith Rabois** (Khosla Ventures) and Felicis, is a strong vote of confidence in both the market opportunity and Runlayer’s execution. Rabois is renowned for supporting transformative technologies at their inflection point, and his backing signals validation of MCP security as a new, high-priority category within enterprise AI[1][2].
## Early Customer Traction: Serving the Unicorns
In an industry where security startups often struggle to gain early trust, Runlayer’s rapid customer acquisition is remarkable. Within just four months, the company has signed eight unicorns and public firms as customers, including:
– **Gusto**
– **Rippling**
– **dbt Labs**
– **Instacart**
– **Opendoor**
– **Ramp**[1][2]
This strong early adoption not only reflects the urgency of the problem, but also demonstrates the practical value Runlayer delivers in real-world, high-stakes environments.
## The Urgency of Securing Autonomous AI
The explosive growth of MCP-based AI agents has enabled unprecedented automation and productivity, but it has also created a new and largely unprotected attack surface for enterprises. As recent incidents have shown, even a single vulnerability in an MCP server can have far-reaching consequences, exposing sensitive data and disrupting business operations[1][2].
Runlayer’s mission — to secure the exploding ecosystem of autonomous AI agents — is therefore more than timely. By combining technical innovation, protocol-level expertise, and deep integration with leading enterprises, Runlayer is positioned to become a foundational layer in the future of secure AI deployment.
## Looking Ahead
Runlayer’s launch reflects a broader trend: as AI agents move from experimental projects to core enterprise infrastructure, **security** becomes a non-negotiable requirement. The company’s comprehensive solution, high-profile backers, and impressive customer base make it a name to watch as the MCP ecosystem matures and the stakes of AI security continue to rise[1][2].
Original source: TechCrunch – MCP AI agent security startup Runlayer launches with 8 unicorns, $11M from Khosla’s Keith Rabois and Felicis