Datadog LLM Observability

Datadog LLM Observability adds tracing, evaluation and security checks for LLM apps and agents to the Datadog monitoring platform. It suits companies already using Datadog that want AI traces next to their infrastructure and application data.

It captures prompts, responses, token use, latency and errors for each step of an agent, and runs built-in quality and safety evaluations such as hallucination and prompt-injection checks. An agents console shows how agents behave across services.

Pricing is usage-based on top of other Datadog products, with a free trial available. Traces link to APM, logs and real user monitoring for end-to-end debugging.