Monte Carlo: Building Data + AI Observability Agents with LangGraph and LangSmith

Data observability platform Monte Carlo built an AI Troubleshooting Agent to help data teams resolve pipeline issues more quickly. The agent uses LangGraph to create a graph-based workflow that investigates multiple potential root causes simultaneously, a significant improvement over manual, sequential troubleshooting. This system can launch hundreds of sub-agents to check for code changes, analyze event timelines, and investigate dependencies in parallel. The team utilized LangSmith for debugging and visualizing the agent's complex workflows, which accelerated development and allowed for rapid iteration on prompts.