Put Your AI Agent Security Policies to the Test with CyberServal AIDR

An approved AI agent can still make an unapproved connection. Would your security team know?

A developer uses an approved coding agent to update a project. They switch its model connection to a third-party relay that the company has not approved. This intermediary forwards requests to a model provider, meaning source code or internal information included in those requests may pass through it.

The agent remains on the approved software list, while the way it handles company information has changed. A software inventory alone may not reveal that difference.

Know Which Services Your Agents Actually Use

CyberServal AIDR (AI Detection and Response) helps security teams discover agents and their connected assets across supported endpoints, including skills, MCP servers and model configurations.

For the scenario above, teams can examine the model services an agent uses and locate configurations that require attention. Risk findings link to supporting evidence and affected endpoints, helping administrators determine where to act.

In supported scenarios, administrators can configure policies to audit or block connections to unapproved model services and relays. Policies can also audit or block protected file reads.Session records and linked audit evidence help teams investigate activity and review the recorded policy outcome.

Built for Enterprise Environments

AIDR supports evaluation across existing enterprise environments through:

• Compatibility with more than 100 mainstream AI agents, helping teams assess the agents they already use.
• Windows and macOS endpoint support across both platforms.
• Millisecond-level local sensitive-content detection, combining rules with local semantic classification.
• On-premises deployment, allowing enterprises to deploy AIDR within their infrastructure and retain platform data on premises.

Supported capabilities depend on the operating system, agent version and integration.

Test AIDR with Your Own Agents and Workflows

CyberServal is inviting 30 eligible enterprises to join its AIDR Enterprise Trial Program, offering a free three-month trial.

For a team concerned about unapproved model connections, the evaluation can focus on a practical question: do your configured policies respond as intended when an agent attempts an unapproved connection? Teams can test supported controls and review the resulting audit evidence against their own requirements.

Participants receive deployment guidance, trial support and a review of evaluation results. Technical compatibility and evaluation scope are confirmed before activation.

Applications close on January 30, 2027, at 23:59 SGT.

Apply for the Enterprise Trial

Participation is subject to eligibility and technical compatibility.

Prefer a guided walkthrough first? Request a Demo