Guard ConsoleA content-safety checkpoint in front of every LLM call
Guard Service puts one gateway endpoint between your applications and your models — flagging PII, jailbreaks, and harmful content before they go anywhere, with a full audit trail behind it.
bolt
Real-time interception
Every prompt is screened by IBM Granite Guardian before it reaches your model. PII, jailbreaks, and harmful content are flagged inline, with a reason attached.
policy
Policy engine
Turn detection categories on or off per tenant and see the effect immediately in the Playground.
monitoring
Global observability
Every invocation, verdict, and token count lands in one auditable log, broken down by tenant, API key, and model.