promptshield-firewall

PromptShield is a real-time API for blocking prompt injection attacks before they reach an LLM. It is described as an LLM prompt injection firewall and is presented as production-ready, with a free tier available for developers and side projects.

The service analyzes each user input and returns a live response with a verdict, risk score, confidence value, explanation, recommendation, latency, and sanitized input field. Its documented verdicts include BLOCKED, SUSPICIOUS, and SAFE. The page says it detects 12+ attack vectors and gives examples such as jailbreaks, prompt extraction, role hijacking, indirect injection, context manipulation, and data exfiltration. It also names DAN, AIM, Developer Mode, and 300+ known jailbreak templates, and shows that blocked requests can be dropped, suspicious ones logged and reviewed, and safe ones forwarded to the LLM.

Integration is shown through a single HTTP request to POST /v1/analyze before the model call, with the API key created by posting to /v1/keys with an email address. The page also provides a Python SDK install command and example code for checking whether a message is safe or scanning it for full details. Swagger docs access is listed on the free plan, and a live demo is available without signup.

Pricing starts with a free tier at $0 per month, described as forever free for developers and side projects, with 50 API calls per month. The Pro plan is $99 per month for startups and production LLM applications and includes 25,000 API calls per month. The Business plan is $499 per month for scaling teams and includes 500,000 API calls per month. An Enterprise plan uses custom pricing and includes unlimited API calls, on-premise deployment, SOC 2 readiness, custom ML model tuning, dedicated support, a 99.99% SLA, and white-label availability. The page footer identifies the company as PromptShield Inc., and the site also mentions a contact name, Kehinde Omosor.