LLM analytics demos & examples

Explore real-world examples of LLM analytics in action. These projects demonstrate different ways to integrate PostHog's LLM analytics features.

Example projects

These are projects some of our team members have built to showcase LLM analytics capabilities and help with development and debugging.

LLM Analytics Apps - Interactive CLI showcasing PostHog LLM analytics across 13+ AI providers (Anthropic, OpenAI, Google, LangChain, and more). Great for understanding how different SDKs integrate with PostHog. We use this mainly for help with dev, debugging and reproduction of user issues etc.
Andy's Daily Factoids - Full-stack AI factoid generator demonstrating LangChain callbacks, multi-platform observability, cost tracking, and real-time streaming. Also uses OpenRouter for the LLM calls.
GitHub Standup Agent - Multi-agent standup summarizer using OpenAI Agents SDK. Demonstrates automatic agent tracing with posthog.ai.openai_agents.instrument().
Repo Stalker - GitHub PR/Issue AI assistant with custom span tracking for different interaction types (chat, summaries, analysis). This is built by Lovable and uses manual capture
Plot Agent - AI data visualization assistant using LangGraph. Shows PostHog callback integration with conversation session tracking.
Anomaly Agent - LLM-powered anomaly detection for time series data using LangGraph. Demonstrates multi-step AI workflows with verification.
HogFlix (PostHog Demo 3000) - Full PostHog feature showcase including LLM chat with both LangChain and direct OpenAI integration patterns.
Anomstack - Anomaly detection system with LLM-powered analysis. Traditional app with product analytics and agent tracing using LLM analytics.