Promptfoo tops GitHub for AI red teaming

AI security is finally getting its own developer-native toolchain, and Promptfoo is one of the clearest breakouts. Its traction comes from collapsing evals, red teaming, and AppSec-style automation into one workflow instead of treating them as separate disciplines.

• –The CLI-first approach matters: teams can run evals, red-team probes, and regression checks locally or in CI without rebuilding their stack around a SaaS dashboard
• –Promptfoo is model-agnostic across GPT, Claude, Gemini, Llama, Azure, Bedrock, and Ollama, which fits the messy multi-model reality most AI teams now operate in
• –The product site now frames Promptfoo as a broader AI security platform with guardrails, code scanning, model security, and MCP tooling, not just a prompt-testing harness
• –Outside comparison pages increasingly place it alongside AI security tools like Garak, Giskard, and Protect AI, which is a good sign the category is maturing into a real market
• –The recent “joining OpenAI” announcement adds even more momentum, but the company says the open-source project will continue, which should keep the repo relevant for practitioners