Workflows | AI Agent Security

AI coding assistants act on your endpoints with the privileges of the person running them. They read files, run commands, reach cloud resources, and call MCP servers, often faster than anyone can review what they did. These workflows describe what AI Agent Security lets you do about that, in the order most organizations approach it: see what is running, control it, then prove what happened.

Workflow	Description
Deploy the agent and build your inventory	Install the AI Agent Security agent and review the AI agents, MCP servers, LLM endpoints, and agent skills it finds.
Map agent privileges to cloud resources in Identity Security Insights	Trace the privileges an AI agent inherits to the cloud resources it can reach in Identity Security Insights.
Block model providers and unapproved agents	Block a model provider, an unapproved tool, or an MCP server you have not approved.
Block access while retaining user privileges	Stop an AI agent from an action while the person who runs it keeps that access.
Review and decide exception requests	Work the approval queue when a policy blocks a legitimate action.
Investigate AI coding sessions and actions	Pull the record of an AI session and follow it down to a single action.