What is Datadog Agent Observability?
Datadog Agent Observability is a platform for evaluating, tracing, and monitoring AI agents from development through production. It connects agent behavior with application, infrastructure, and end-user experience signals in the Datadog platform.
Key Capabilities
- Agent tracing: Captures prompts, retrieval steps, tool calls, agent decisions, latency, token usage, retries, and errors.
- Experiments: Compares prompts, models, and agent configurations against shared datasets before release.
- Evaluation: Provides built-in and custom evaluators, human review, annotations, and versioned datasets created from production traces.
- Production monitoring: Tracks agent quality, cost, latency, and reliability, with alerts for operational issues and regressions.
- Security and governance: Helps detect unsafe outputs, prompt-injection attempts, hallucinations, and personally identifiable information (PII) exposure.
- Full-stack correlation: Links LLM traces with application performance monitoring, infrastructure metrics, service dependencies, and real-user monitoring sessions.
Audience & Use Cases
- Audience: AI engineers, data scientists, SRE and DevOps teams, and engineering leaders operating AI-enabled applications.
- Use cases: Validating agent changes before deployment, diagnosing production failures, monitoring cost and latency, creating regression datasets, and understanding AI-related user impact.
Technical Specifications
- Integrations: Supports major model providers and agent frameworks, including OpenAI, Anthropic, Google Gemini, Amazon Bedrock, LangChain, Pydantic AI, and Vercel AI SDK.
- Platform integration: Uses Datadog tracing and monitoring capabilities across development and production workflows.
Categories & Use Cases
Technical Details
| Mobile Application | No |
|---|
FAQs
What is Datadog Agent Observability?
Datadog Agent Observability is a platform for evaluating, tracing, and monitoring AI agents from development through production. It connects agent behavior with application, infrastructure, and end-user experience signals in the Datadog platform.