LLM Gateways

LLM Gateways is an AI security layer for LLM applications. It is built to stop prompt attacks before they reach a model, including prompt injection, jailbreaks, and system-prompt extraction. The service presents itself as a single API call with instant risk scores, blocking malicious prompts in milliseconds while also returning analytics and webhook notifications.

The product supports a drop-in SDK or raw HTTP access. Examples are shown for Python, Node.js/TypeScript, and cURL, and the Python and Node.js examples wrap existing OpenAI clients. The request flow returns a risk score, an action such as block, and a list of threats; the page also shows a blocked-request exception in the SDK examples. Its detection stack is described as a three-layer engine made up of pattern rules, semantic similarity, and an LLM judge.

LLM Gateways says it covers 78+ threat patterns and lists prompt injection, DAN jailbreaks, system-prompt extraction, PII leakage, and token smuggling among them. It also offers real-time dashboards, threat breakdowns by category, per-API-key statistics, scan history, HMAC-signed webhooks with configurable event types, email alerts when block rates spike, and per-key custom rules for blocklisting phrases or disabling detection categories. Each API key is rate limited to 100 requests per minute.

The service is available free during beta, with all features included and no credit card required. The page identifies Apostrophe AI as the publisher and links to docs, privacy, and terms.