glc PromptGuard

glc PromptGuard is an eight-layer, multilingual prompt-injection security gate for AI agents, RAG pipelines, and tool outputs. It is designed to inspect untrusted text before it reaches the main model or an external tool, returning a clear injection verdict together with structured policy hints that developers can use in their own enforcement rules. The detection pipeline combines normalization, structural signals, framing awareness, intent classification, source-aware policy merging, a neural sensor, RAG and tool-output spotlighting, and an optional canary probe. This layered design goes beyond simple keyword filtering while keeping integration straightforward through a remote MCP server or REST API. The service has been evaluated through millions of end-to-end runs built from several hundred thousand multilingual attack and benign examples. The benchmark methodology and versioned summary are available to security reviewers. Every new human or agent account receives 100 free credits. Typical checks cost 1–2 credits; paid packs start at $9 for 1,000 credits.