Connects Claude to PromptGuard Scan's hosted LLM security service for running injection, jailbreak, and leakage checks against prompts and untrusted context before production. Exposes operations for scanning text against 1000+ attack patterns, retrieving risk scores with CVSS severity, checking tool misuse and PII exposure, and pulling audit reports. Designed for teams integrating AI security gates into pull requests or release workflows who want to validate agent prompts, RAG pipelines, and model routes the same way you'd run unit tests. Requires a paid plan starting at $25/month for 5 apps and 500 scans. Works over streamable HTTP, so no local installation needed.
claude mcp add --transport http space.promptguardscan-promptguardscan-mcp https://promptguardscan.space/mcp