Latitude provides error monitoring and observability for AI agents, functioning like Sentry specifically for LLMs. It automatically groups failed traces into trackable issues, tracks evaluation alignment over time, and indexes all traces for semantic search. The platform supports multi-turn agent sessions with full execution paths and tool call visibility.