CyberArk launches new AI security tools on AWS Marketplace

CyberArk has announced the availability of its Secure Cloud Access (SCA) MCP Server and Agent Guard tools in the AWS Marketplace AI Agents and Tools category.

With this development, customers can now use AWS Marketplace to discover, purchase, and deploy AI agent security solutions directly through their AWS accounts, a move aimed at further accelerating the deployment of agentic AI workflows with enhanced security measures.

The SCA MCP Server and Agent Guard tools are designed as part of the CyberArk Identity Security Platform, introduced to address increasing enterprise concerns about the privileged access of AI agents and to promote identity-first security practices at scale. The company cites internal research indicating that 68 percent of organisations currently lack identity security controls for AI, resulting in challenges for security teams attempting to maintain oversight in complex, multi-cloud environments.

According to CyberArk, Secure Cloud Access MCP Server integrates Zero Standing Privileges (ZSP) directly into development workflows, aiming to reduce the risks associated with persistent entitlements, credential sprawl, and unmanaged AI agents. Meanwhile, Agent Guard, which is also available as an open-source tool, allows developers to integrate their AI agents with credential management systems such as AWS Secrets Manager and CyberArk Secrets Manager, offering additional layers of credential protection.

"The promise of agentic AI can be undermined by inadequate security controls, which introduce risk and increase the likelihood of a breach," said Peretz Regev, Chief Product Officer at CyberArk. "With SCA MCP Server and Agent Guard, organisations can implement Zero Standing Privileges to maintain more secure and scalable AI-first operations, helping to stop excessive standing access, privileges and permissions from becoming scattered through cloud environments. By offering them through AWS Marketplace, organisations will have greater access to these critical tools."

The SCA MCP Server enforces Zero Standing Privileges across multi-cloud set-ups, and provides scoped access for AI assistants like Amazon Q and Anthropic's Claude, offering audit trails and supporting role-based access controls. Developers are able to efficiently request secure, privileged access for AI assistants directly from their integrated development environments or command line interfaces. The product is designed to apply least privilege principles not only to human users, but also to machine identities throughout CI/CD pipelines and agentic AI workflows.

Agent Guard offers real-time monitoring of large language model (LLM) and tool calls, generating analytics-ready graphs, and logging tool input and performance data to aid with metadata capture across a variety of AI frameworks. The tool supports multiple secret providers, including CyberArk Secrets Manager and AWS Secrets Manager, and ensures that API keys and related secrets are provisioned as environment variables at runtime.

By having these tools available in the AWS Marketplace AI Agents and Tools category, CyberArk is aiming to streamline the procurement process for AI security, reducing the need for time-consuming vendor assessments and complex negotiations. Through centralised purchasing with AWS accounts, clients can maintain oversight of their licensing, payments, and user access from within their existing AWS infrastructure.

CyberArk states that these measures align with its broader commitment to identity security for human and machine identities within modern enterprise environments. The move comes as organisations are under increased pressure to manage privileges and credentials in software-driven, multi-cloud operations, and to address security gaps that might expose sensitive access via unmanaged or poorly secured AI agents.

The features provided by SCA MCP Server and Agent Guard are aimed at organisations looking to bolster their security posture as they adopt AI-driven automation, workflow agents, and related technologies within their technical and operational environments.