Galileo

Galileo is an AI observability and evaluation platform for teams building generative AI applications and agents. It brings development datasets, production traces, evaluator results, and reviewer feedback into one workflow so teams can find failure patterns and decide what to fix. It supports more than 20 ready-made evaluations for RAG, agents, safety, and security, alongside custom evaluators for domain-specific requirements. Reviewers can correct scores and use Autotune to revise evaluator prompts; Galileo reports that five corrections can improve evaluator accuracy by up to 17% in demonstrated cases. Optimized evaluations can be distilled into Luna models and used as production guardrails, with the company stating that this approach can monitor all traffic at 96% lower cost. The free plan is practical for experiments with 5,000 traces each month, unlimited users, and unlimited custom evaluations. Pro starts at $100 per month when billed yearly and raises the allowance to 50,000 traces, while usage-based scaling can increase the price. Enterprise buyers can request hosted, VPC, or on-premises deployment, stronger access controls, dedicated support, and unlimited traces. Galileo suits teams that want evaluation, debugging, monitoring, and guardrails in one platform, but adopters should check trace volume, deployment requirements, and support needs before choosing a paid tier.