Welcome to Circuit Breaker Labs CLI
The Circuit Breaker Labs CLI is a powerful command-line tool for evaluating the safety of AI language models. Test your models against adversarial prompts, measure safety thresholds, and ensure your AI systems respond appropriately to harmful content.Quick Start
Get up and running with your first evaluation in minutes
Installation
Install the CLI for Linux, Mac, or Windows
Command Reference
Explore all available commands and options
Custom Providers
Integrate custom model endpoints with Rhai scripting
Key Features
Single & Multi-Turn Evaluations
Test models with both single-turn prompts and multi-turn conversations
Multiple Providers
Support for OpenAI, Ollama, and custom model endpoints
Interactive TUI
Real-time progress visualization with an interactive terminal interface
Configurable Thresholds
Set custom safety score thresholds for your use case
Custom Scripting
Integrate any API using Rhai scripting language
JSON Output
Export detailed evaluation results in JSON format
How It Works
The Circuit Breaker Labs CLI connects to the Circuit Breaker Labs API to retrieve adversarial test cases, sends them to your language model, and evaluates the responses for safety concerns.Choose Evaluation Type
Run single-turn evaluations for quick tests or multi-turn for conversational safety
Get Started
Ready to evaluate your model?
Follow our quickstart guide to run your first evaluation