AgentGuard FAQs

Question 1

Are AgentGuard's decisions and agent actions auditable?

Accepted Answer

Yes, AgentGuard provides cryptographically auditable decisions for every action. It maintains a persistent audit trail that includes caller identity and HMAC signatures, ensuring transparency, non-repudiation, and comprehensive forensic capabilities for all AI agent activities.

Question 2

How does AgentGuard handle escalating security risks?

Accepted Answer

It uses a persistent, escalation-only agent state model (active, monitoring, approval_required, suspended, killed). Based on detected risks, an agent's state automatically escalates, triggering enhanced auditing, requiring human approval, or completely blocking its operations until the risk is mitigated or resolved.

Question 3

What is AgentGuard and what problem does it solve?

Accepted Answer

AgentGuard is a policy-driven runtime security and governance layer specifically designed for AI agents. It addresses the challenge of controlling agent behavior by enforcing policies across four distinct stages: pre-execution, payload, advanced governance, and post-execution, ensuring agents operate securely and compliantly.

Question 4

What kind of policy-driven controls does AgentGuard offer?

Accepted Answer

AgentGuard allows you to define and enforce various policies, such as rate limits, spend limits, secret exposure checks, payment policy enforcement, cross-tool anomaly detection, and role-based access control (RBAC) to precisely govern agent behavior and resource usage.

Question 5

How does AgentGuard secure AI agents throughout their lifecycle?

Accepted Answer

AgentGuard employs a comprehensive four-layer security architecture with 20 specialized tools. This includes pre-execution gates (e.g., risk scoring, rate limiting), payload security (e.g., secret exposure, prompt injection checks), advanced governance (e.g., RBAC, anomaly detection), and post-execution output scanning for PII, secrets, and exfiltration.

AgentGuard

AgentGuard

Key Features

Use Cases

Key Features

Use Cases