Agent Guardrails protect both sides of an agent interaction.

Guardrail	Control point
Context Guardrails	Checks external content before an agent consumes it
Runtime Guardrails	Monitors or blocks risky actions where a coding agent runs

Both guardrails send findings into the security factory. Customer agents and workflows act on those findings, structured agent reports return through Superagent MCP, and Superagent retains the reports as context for later work.

Next steps

• Check external content with Context Guardrails
• Monitor agent actions with Runtime Guardrails
• Write and enforce runtime rules