Promptfoo

Founded by Ian W. in 2023

Detect prompt injections and security vulnerabilities in your AI applications

Pricing model

Free Trial

Rating

★ People love it

Time to value

Quick Setup (< 1 hour)

You can use Promptfoo to test AI applications, agents, and retrieval-augmented systems for security vulnerabilities. The CLI runs automated red teaming to simulate real user attacks, finding direct prompt injections, jailbreaks, data leaks, and toxic content before deployment. You can define test cases in simple configuration files, run scans directly inside your pull requests and continuous integration pipelines, and compare model outputs across multiple foundation providers. It provides clear remediation advice so your engineering and security teams can fix prompt weaknesses quickly.

What Promptfoo does

Configure prompt evaluation suites in YAML filesSimulate automated adversarial attacks against LLM endpointsReview security findings directly within pull requestsScan agent tools for unauthorized execution risksCompare output quality across different foundation modelsEnforce safety guardrails against prompt injection attemptsAutomated red teaming for AI agents and RAG setupsVulnerability scanning for direct and indirect prompt injectionsJailbreak testing tailored to custom guardrailsCI/CD pipeline integration with security feedback in pull requestsDetection of sensitive personal data leaks in responsesLocal evaluation runs configured through simple YAML files