Question 1

What is AI agent governance?

Accepted Answer

AI agent governance is the practice of managing identity, policy, and audit for autonomous AI agents. It ensures that every agent is identified, constrained by policies, and its actions are logged for compliance. MeshGuard provides a runtime control plane that enforces governance at the point of agent action.

Question 2

How does MeshGuard work?

Accepted Answer

MeshGuard sits between your AI agent orchestrators (LangChain, CrewAI, OpenAI Agents, etc.) and the tools and data they access. Every agent action passes through MeshGuard's gateway, which validates the agent's identity via JWT credentials, evaluates YAML policies, checks trust scores, and logs the decision to an immutable audit trail — all in real time.

Question 3

What AI agent frameworks does MeshGuard support?

Accepted Answer

MeshGuard supports LangChain, CrewAI, AutoGPT, OpenAI Agents SDK, Claude Code, AWS Bedrock, Google Vertex AI, Microsoft Semantic Kernel, and any agent framework via its generic REST API. SDKs are available for Python, JavaScript/TypeScript, and .NET.

Question 4

How is MeshGuard different from OPA or LangChain Guardrails?

Accepted Answer

OPA is a general-purpose policy engine; MeshGuard is purpose-built for AI agents with trust tiers, delegation chains, anomaly detection, and agent-specific identity. LangChain Guardrails are framework-specific content filters; MeshGuard provides external governance that works across all agent frameworks and enforces identity, policy, and audit at runtime.

Question 5

What is agent trust scoring?

Accepted Answer

MeshGuard's trust scoring system assigns each agent a 0-100 score based on five weighted components: behavioral history (20%), anomaly patterns (30%), delegation usage (20%), tenure (15%), and administrator vouchers (15%). Agents are placed in trust tiers (Unverified, Verified, Trusted, Privileged) that determine their policy access level.

Question 6

Can MeshGuard detect prompt injection attacks?

Accepted Answer

Yes. MeshGuard includes real-time streaming middleware that detects prompt injection attempts, data exfiltration patterns, and other adversarial inputs before they reach your agents or downstream systems. It operates at the gateway level, protecting all agents regardless of framework.