Open source dashboard for AI engineering & LLM data
OpenLIT provides zero-code observability for AI agents and LLM apps. Monitor your full stack, from LLMs and VectorDBs to GPUs, without changing any code. See exactly what your AI agents are doing at every step. Catch problems before they reach users, improve response quality over time, and optimize costs as you scale. Ship reliable AI faster with complete visibility and control. It's OpenTelemetry native, open source, self-hostable, and works with 50+ providers.
OpenLIT is OpenTelemetry-native and exports standard OTLP data. Configure any OTLP-compatible backend (Grafana, Datadog, New Relic, SigNoz) as the destination, avoiding vendor lock-in and enabling integration with existing monitoring stacks.
OpenLIT includes Prompt Hub for centralized prompt versioning and management, OpenGround playground for side-by-side model comparison, and AI Model Evaluation tools to run online/offline evaluations via UI or SDK to find best-performing prompts and models.
OpenLIT uses automatic instrumentation via monkey-patching; existing OpenAI, Anthropic, or LangChain code requires no modification. Kubernetes Operator enables zero-code observability injection into containerized workloads without rebuilds.
OpenLIT auto-calculates costs from model pricing, tracks input/output tokens, and provides unified dashboards across 60+ LLM providers (OpenAI, Anthropic, Bedrock, etc.). Fleet Hub enables cost comparison across multiple deployments and environments.
OpenLIT provides distributed tracing with OpenTelemetry to visualize request flows, identify bottlenecks, and understand the complete lifecycle of AI interactions. Automatically captures token usage, latency, cost, and model details with one line of code (openlit.init()).