Prompt Leak Guard
Detect, redact, and block credentials, PII, payment data, and other sensitive values before they reach an AI model, MCP tool, or log.
How to pay
Subscribe
$19/month
Predictable monthly cost with included usage. Best for steady, high-volume traffic.
- Unlimited tools within plan limits
- One API key, billed once a month
- Cancel any time
Stop sensitive data before it reaches a model, tool, log, or third-party service.
Prompt Leak Guard is a deterministic, stateless safety gate for AI agents and MCP workflows. It scans plain text and nested JSON for common credentials such as AWS access keys, GitHub and Slack tokens, bearer tokens, JWTs, private keys, and generic API keys. It also detects financial and identity data such as Luhn-validated payment cards, mod-97-validated IBAN candidates, email addresses, phone numbers, US Social Security numbers, IP addresses, and URLs containing embedded credentials.
Use four focused tools to protect a workflow:
- scan_text identifies sensitive values with category, severity, confidence, and safe character offsets.
- redact_text replaces findings with labels, masks, hashes, or a custom marker.
- scan_json recursively inspects keys and values while reporting exact JSON paths.
- evaluate_policy produces an explainable allow, review, or block decision from your own category, severity, and finding-count rules.
Raw matches are never returned in finding objects. Payloads are processed in memory, are not stored by the application, and are never written to application logs. The server requires no API keys, database, or external data provider and makes no outbound API requests.
Prompt Leak Guard is useful as a preflight check before sending customer messages to an LLM, passing MCP tool arguments, persisting agent traces, or forwarding automation payloads to another service. It is a detection and redaction aid, not a certification of GDPR, HIPAA, PCI DSS, or any other compliance standard.