AI Safety & LLM Guardrail APIs
Detect prompt injection, redact reversible PII placeholders before LLM prompts leave your VPC, catch RAG hallucinations, and scan LLM outputs in sub-millisecond edge compute.
Zero-retention Cloudflare Worker APIs organized into seven developer domains.
Detect prompt injection, redact reversible PII placeholders before LLM prompts leave your VPC, catch RAG hallucinations, and scan LLM outputs in sub-millisecond edge compute.
Audit Model Context Protocol (MCP) manifests for tool poisoning, validate LLM tool-call arguments, inspect shell/SQL commands before execution, and sign tamper-evident agent action receipts.
Repair broken LLM JSON output, convert noisy web pages to Markdown, normalize multi-turn chat messages[], chunk docs with code/table boundaries, and neutralize CSV formula injection.
Block private IPs and cloud metadata SSRF attacks, mask 28 secret types in logs, audit HTTP security headers, verify Stripe/GitHub/Polar webhooks, and validate RFC 8725 JWT/JWKs.
Audit MX, SPF 10-lookup limits, DKIM selectors, DMARC policies, and MTA-STS, generate valid DNS TXT records, and issue RFC 8058 stateless one-click unsubscribe consent tokens.
Validate EU VAT numbers via VIES, MOD-97 IBAN/BIC codes, LEI/EORI corporate IDs, EN 16931 / Peppol BIS 3.0 UBL invoices, Hungarian NAV rules, EU workdays, and global sanctions watchlists.
Detect breaking JSON Schema/OpenAPI diffs, parse cron schedules with human explanations, inspect AI crawler bots and robots/llms.txt files, strip UTM tracking, and prevent ReDoS.