Agent classes / Eval and judge agent
Eval and judge agent Infrastructure Kit
Scores a trace or a patch against a rubric, without the gold label in its prompt, and records the harness version.
Prompt/agent eval harness
Depot Pick: Braintrust
OSS / CI-native LLM evals
Depot Pick: Promptfoo
RAG-specific evaluation metrics
Depot Pick: Ragas
Production guardrails & hallucination mo…
Depot Pick: Galileo
Eval CI gates
Depot Pick: Braintrust
Red-team evals
Depot Pick: Promptfoo
Synthetic eval data
Depot Pick: OpenAI
LLM/agent tracing
Depot Pick: Langfuse