Loading tool details…
Datadog LLM Observability (part of Datadog's Agent Observability suite) is the enterprise APM's answer for AI agents and LLM apps: end-to-end tracing of prompts, tool calls, and retrieval steps; quality evaluations; cost tracking per model; and security detections — all inside the same UI where your infra, APM, and RUM traces already live.
Datadog LLM Observability (part of Datadog's Agent Observability suite) is the enterprise APM's answer for AI agents and LLM apps: end-to-end tracing of prompts, tool calls, and retrieval steps; quality evaluations; cost tracking per model; and security detections — all inside the same UI where your infra, APM, and RUM traces already live.
Datadog LLM Observability instruments AI agents and LLM applications with the same tracing model Datadog already uses for microservices. You install the SDK (Python, Node, .NET, Go, Java) or drop in an OpenTelemetry-based integration for /tools/langchain, /tools/langgraph, /tools/langfuse-tracked apps, LangSmith exports, or /tools/crewai; every prompt, model call, tool invocation, retrieval step, and downstream API call appears as a span in a distributed trace. The product bundles model-quality evaluations (hallucination, toxicity, prompt injection, PII leakage), cost tracking that rolls up token spend by team/service/model, and prompt clustering that surfaces the most expensive or most-failed intents.
Unique to Datadog: LLM traces stitch directly into APM traces, so you can follow a request from the user's browser (RUM), through your API (APM), into an LLM agent (LLM Observability), out to a downstream vector search or database (Database Monitoring), and back — one trace, one price. Datadog also ships an 'Agent Observability' sub-product for autonomous agent runs with additional views for planner steps, cross-session memory, and multi-agent workflows.
Pricing is not published as a flat SaaS: LLM Observability is priced per session/trace with retention tiers similar to APM Indexed Spans, and typically sold inside an enterprise Datadog contract. Small teams should treat it as an add-on to an existing Datadog seat, not a standalone tool. If you're not already on Datadog, dedicated LLM observability tools like /tools/langfuse (open source), /tools/langsmith, /tools/braintrust, or /tools/helicone are usually faster to adopt and cheaper for LLM-only shops.
Datadog LLM Observability is the natural choice for teams already invested in the Datadog ecosystem. The cross-correlation between LLM performance and infrastructure metrics is genuinely useful for production debugging. However, span-based pricing and auto-activation behavior require careful cost management, and it's overkill if you don't already use Datadog.
Per session/trace (contact sales)
Add-on (contact sales)
Custom
Ready to get started with Datadog LLM Observability?
View Pricing Options →We believe in transparent reviews. Here's what Datadog LLM Observability doesn't handle well:
Datadog has published its State of AI Engineering 2026 report drawing on aggregated production telemetry across thousands of customers, and continues to expand agentic workflow tracing and evaluation coverage for multi-agent systems. Recent platform investments emphasize deeper integration between LLM Observability, Cloud SIEM, and Sensitive Data Scanner to address production safety concerns around prompt injection and data exfiltration in agentic applications.
LLM Observability
Open-source LLM engineering platform that captures traces, runs evaluations, and helps teams analyze cost, latency, and quality across production AI apps.
AI Monitoring
Open-source LLM observability and gateway platform offering one-line integration, request logging, prompt management, evals, and a multi-provider proxy.
AI Observability
Arize Phoenix is an open-source AI observability and evaluation platform for tracing, testing, and improving LLM applications and agents.
LLM Observability
LangSmith helps developers trace, evaluate, test, and monitor applications built with language models and agent workflows.
Pricing and features change often. If anything on this Datadog LLM Observability page is wrong or missing, tell us and include a link to the source. We review every correction.