Promptfoo

What is Promptfoo?

promptfoo is your go-to tool for securing and testing LLM (Large Language Model) applications. Trusted by over 60,000 developers, it helps you eliminate risks, catch regressions, and maximize output quality in your AI projects. Whether you're dealing with PII leaks, prompt injections, or harmful content, promptfoo has got you covered.

What are the features of Promptfoo?

• Automated Red Teaming

  Run custom scans to detect security, legal, and brand risks.
• Comprehensive Security Coverage

  Identify PII leaks, insecure tool use, and cross-session data leaks.
• Developer-Friendly

  Command-line interface, live reloads, and caching for quick iterations.
• Open-Source & Battle-Tested

  Used by teams serving millions of users and supported by an active community.
• Easy Configurations

  Simple declarative config for complex LLM testing.

What are the use cases of Promptfoo?

• Pre-deployment security scans

  for LLM applications.
• Continuous monitoring

  to catch regressions and vulnerabilities.
• Custom probes

  for identifying specific failures in your application.
• Quality evaluations

  to ensure your LLM outputs are reliable and secure.

How to use Promptfoo?

1. Install promptfoo using the command: npx promptfoo@latest redteam init.
2. Define your test cases in a YAML config file.
3. Run custom scans to detect vulnerabilities and improve your LLM's performance.