Silverfort has launched
AI Agent Security, a new solution designed to control and monitor how AI agents access enterprise systems, especially in environments using the Model Communication Protocol (MCP). The product treats AI agents as identities, tying them to human owners, enforcing real-time access controls, and preventing overreach or misuse.
AI Agents and MCP: Why Identity Governance Matters
AI agents operating over the Model Communication Protocol (MCP) can access internal systems dynamically, without always having clear boundaries. That makes them a risk. Without visibility and control, agents can escalate privileges, access unintended resources, or leak sensitive data.
“AI agents need to be secured just like any human user,” said
Yaron Kassner, CTO and co-founder of Silverfort. “Our method is to attach each agent to a human, so it only has access to what it needs and nothing more. By sitting in line between the agent and the MCP server, every action is traced, and we can automatically discover, classify, and monitor AI agent identities.”
Inline Controls Without Slowing Down Agents
Silverfort applies the same identity-first model it uses for human and service accounts to AI agents. Each one is tied to a human owner, monitored in real time, and governed by dynamic access policies.
“Because all of this is centralized and visible in real time,” Kassner explained, “identity teams and CISOs stay in control. With an identity-first approach, you can put governance and guardrails in place, enforce strong access policies around autonomous agents, and pull humans into the loop when needed.”
Performance is built in. AI agents are expected to move fast, so controls can’t add friction.
“We built Silverfort’s AI Agent Security to enforce inline controls without disrupting agent workflows or breaking access to the systems they need,” Kassner said. “Every call is analyzed and logged, access is granted or denied based on dynamic policies, and AI agents are restricted to the exact permissions they’ve been assigned.”
What Comes Next for MSSPs and Partners
Silverfort is starting with direct enterprise deployments but plans to bring partners into the fold.
“At this point, we're working directly with enterprise customers,” Kassner noted. “However, we are in active conversations with our channel partners to add it to their offerings.”
The company also sees room for MSSPs to include this as part of a broader AI governance strategy.
“In the future, we expect products like Silverfort's AI Agent Security to be an integral part of any MSSPs set of solutions,” Kassner said. “At this point, the market is so young, and enterprises are still figuring out their AI agent strategy; they are just scratching the surface on how to secure them.”
As AI adoption picks up speed, identity-based guardrails will be critical to keeping agent activity aligned with enterprise security priorities.
