PromptSentry

PromptSentry is a web app that sits between your product and any LLM API to detect and block risky prompts and responses in real time. It flags prompt-injection attempts, secrets exfiltration patterns, policy violations, and sensitive data (PII/PHI) before they hit your logs or users. You get a lightweight proxy endpoint, configurable rules, and an audit trail showing what was blocked and why. It also offers red-team replay: upload past conversations and run them against updated detectors to see what would have slipped through. This is not a magic “secure AI” button—attackers evolve and false positives happen—but it meaningfully reduces obvious, costly failures for teams shipping LLM features without a dedicated security group. Pricing can be usage-based per 1,000 requests with a small monthly minimum for startups.