Top 5 AI Agent Eval Tools After Promptfoo's Exit

Dev.to / 3/16/2026

📰 NewsSignals & Early TrendsTools & Practical UsageIndustry & Market Moves

共有:

Key Points

OpenAI acquired Promptfoo for $86 million on March 9, raising questions about whether Promptfoo will stay vendor-neutral.
The article highlights five independent alternatives to Promptfoo—DeepEval, Braintrust, Arize Phoenix, LangSmith, and Comet Opik—none of which are owned by a model provider.
DeepEval is an open-source pytest-native evaluation framework with agent metrics like DAG and tool-call, offering no production monitoring and a local self-host option.
Braintrust is a hosted platform offering CI/CD gates, custom evaluation through 8 RAG, production monitoring via traces and scoring, and enterprise-only self-host.
Arize Phoenix blends OSS and cloud with OTEL traces and a free self-host option; LangSmith provides cloud plus self-host with LangChain-native support (enterprise self-host tier); Comet Opik is OSS plus cloud with high-volume traces and an Apache 2.0 license.

Continue reading this article on the original site.