Overview

Organizations are embedding generative AI into their products and workflows at scale, leveraging LLMs like Anthropic Claude to build intelligent applications, automate reasoning tasks, and deliver personalized user experiences. As adoption grows, understanding model performance, usage patterns, and operational costs becomes essential for teams running Claude in production.

New Relic gives you full observability into your Anthropic-powered applications — from raw request metrics to end-to-end distributed traces — so you can proactively identify issues and optimize performance.

Track service health and performance

• Monitor request volume, error rates, and response durations across Claude models in real time.
• Identify latency spikes and throughput degradation before they impact users.
• Track time to first token (TTFT) to measure model responsiveness.

Control cost and quality

• Break down token usage and estimated cost per model and request.
• Validate model behavior and response patterns to catch quality regressions early.
• Set alerts on cost thresholds to avoid unexpected billing surprises.

Debug faster with end-to-end tracing

• Trace prompt flows from initial request through to final response across your entire stack.
• Correlate LLM latency with upstream and downstream service performance.
• Pinpoint failures in prompts, token limits, or third-party integrations with precision.

Get started!

Instrument your application with OpenLLMetry — an OpenTelemetry-based SDK for LLMs — and connect it to New Relic to start monitoring your Anthropic Claude usage. Check out the documentation below to get set up quickly.

More info

Check out the documentation to learn more about New Relic monitoring for Anthropic Claude models.