Stop prompt threats before they reach your LLM
Pre-LLM threat scoring across 25 categories — prompt injection, jailbreaks, PII, malware, and more. <1ms, $0 infra, runs at the edge (Cloudflare Workers / WASM) and natively in-process.
User Prompt → Sprappy Filter → [Allow / Sanitize / Block] → LLM
Try it live
25 threat categories
Injection, jailbreak, PII, secrets, malware, CSAM, self-harm, and more — one pass.
<1ms, $0 infra
Aho-Corasick over a 57K-pattern DB in a single O(n) pass. No model call, no GPU.
Edge + native
Runs as a Cloudflare Worker / WASM at the edge, or in-process via FFI.