LiteLLM Lens
Once your agents are in production, you cannot manually review every trace.
LiteLLM Lens uses AI agents to analyze your agent traces and find recurring problems. You specify the expected behavior. Lens investigates failures, groups similar problems, and links each finding to the original traces.
Use Lens > Traces to manually inspect individual runs. Use Lens > Investigations to investigate a set of runs, on demand or on a schedule.
Before setup, click Preview sample beside the Lens title to explore sample traces, investigations, and linked evidence. Exit demo returns to your own workspace.
Get started
Deploy Lens, then send your first trace. Choose an integration from the sidebar for a runnable project.
Use the coding agent guide to record personal Claude Code or Codex sessions. Once traces are available, run an investigation or use the API reference.
