Skip to main content

Integration Overview

To ensure the safety and reliability of autonomous workflows, our platform provides a dual-layered approach to AI Agents Security. We offer flexible integration paths designed to fit seamlessly into your existing infrastructure, whether you are using low-code orchestration environments or custom-built cloud agents.

Our goal is to provide a comprehensive security layer that provides visibility into your agent ecosystem and ensures every agent action remains within its intended scope.

The Three Pillars of AI Agents Security​

1. AI Agents Posture Management​

We provide continuous oversight of your agent landscape through automated discovery and risk assessment. This allows security teams to identify vulnerabilities in agent configurations, monitor access permissions, and assess the potential impact of an agent’s capabilities to mitigate risks before a real attack happens.

2. Detection & Response​

We detect suspicious or unintended agent behavior and support investigation workflows by analyzing agent activity signals (for example, passive log-based detection). This provides continuous monitoring and alerting for further investigation without requiring the actual enforcement hooks.

3. Runtime Protection​

This serves as your last line of defense. By monitoring agents during active sessions, we make sure that the agent is doing only what it’s intended to do. We evaluate every tool execution request against your specific security rules, preventing unauthorized actions or unintended behaviors in real-time.

Integration Paths​

Depending on how your agents are developed and hosted, we offer two primary integration methods:

Path A: AI Agents Creation Studios​

Target Platforms: Microsoft Copilot Studio, ServiceNow Agent Studio, n8n, Workato, and similar platforms.

For organizations using specialized creation environments, we integrate directly with the platform’s management and execution layers.

  • AI Agents Posture Management: Accomplished via a Read-Only API connection to the studio. This allows us to map all existing agents and their configurations without disrupting the development lifecycle.
  • Detection & Response: Achieved via platform telemetry (e.g., audit logs and/or execution events) to identify suspicious behavior and deviations from expected agent intent.
  • Runtime Protection: The implementation varies between platforms; each has its own architectural approach. Where supported, this is typically achieved via webhooks triggered during tool execution, ensuring that every action is validated against our security engine.

Path B: Custom & Cloud-Deployed Agents​

Target Environments: Agents deployed on AWS, Azure, GCP, or private clouds via frameworks like LangChain, CrewAI, or proprietary code.

For custom agents, we utilize an LLM Gateway as an integration point to provide our security capabilities.

  • AI Agents Posture Management: By leveraging the LLM Gateway, we gain a unified view of all agents interacting with LLMs, helping the organization understand which agents are active and how they are configured.
  • Detection & Response: We leverage model invocation logs to detect and alert on suspicious or unintended behavior (even in environments where active enforcement is not possible).
  • Runtime Protection: We use the Gateway to gain full visibility into prompts and tool executions, enabling us to enforce our unique security rules on each agent individually.

LLM Gateway & Deployment Support​

Our platform is built to work with your existing architecture. We support several ways to utilize a gateway for securing custom agents:

  • Existing Infrastructure: We integrate with popular solutions like Microsoft AI Gateway (via APIM) and LiteLLM.
  • New Deployments: If your organization does not currently utilize an LLM Gateway, we can provide our own solution or help with deploying one in the organization.

If you require guidance on selecting the right integration path or need help utilizing an LLM Gateway in your organization, please reach out to our technical team.

Note​

Custom Integrations: If you are using a gateway or platform not listed here, please tell us and we will add support in no time.