Claude Code observability
Trace every Claude Code session end-to-end. Capture LLM calls, tool calls, cost, and latency without any proxy — debug autonomous agent runs from one dashboard.
Trace every Claude Code run
Full span traces
See the complete execution tree of every Claude Code run — nested spans, tool calls, and LLM requests on one timeline.
Cost & latency analytics
Track token usage, cost, and latency for every Claude Code call, broken down by model, prompt, or metadata.
Rich metadata & search
Tag, score, and search every request. Filter production traffic by content, model, status, or custom key-value pairs.
OpenTelemetry-native
Claude Code streams traces straight to PromptLayer's OTLP endpoint — no proxy in your request path, no SDK rewrite.
Debug in the playground
Open any Claude Code trace in the Playground to reproduce, tweak, and fix the exact prompt that failed.
Turn traces into evals
Promote real Claude Code runs into versioned datasets and run evaluation pipelines to catch regressions.
Understand what your Claude Code app is doing
Claude Code works autonomously across dozens of steps. PromptLayer records each session as a structured trace, so you can see exactly what the agent read, wrote, and decided.
See the full picture
Every Claude Code run becomes a searchable, replayable trace — inputs, outputs, models, and timing.
Find the bottleneck
Pinpoint the slow span or expensive model call dragging down your Claude Code pipeline.
Catch failures fast
Surface errors, failed tool calls, and low-quality outputs before your users do.
Ship with confidence
Connect traces to evaluation pipelines so every change to your Claude Code app is tested.
Frequently asked questions
If you still have questions feel free to contact us at sales@promptlayer.com