Lunary

Lunary is an LLM observability and prompt management platform for teams building AI agents and chatbots. It helps developers investigate production errors and review how people interact with their applications.

• Agent tracing: Search prompts, results and error stack traces to investigate agent behavior, with alerts when performance falls short.
• Chatbot analytics: Replay conversations and track model usage, costs, frequent topics and user satisfaction through custom dashboards.
• Prompt collaboration: Share templates with technical and non-technical teammates, track versions, run A/B tests and try prompts with different models in a playground.
• Response review: Combine human reviews, response scores and user feedback to assess quality, and label interaction data for fine-tuning.
• Multimodal monitoring: Capture interactions involving text, audio and images, with integrations including OpenAI and LangChain.

You can self-host Lunary on your own infrastructure through Docker or Kubernetes, including inside a VPC. Its separate paid cloud service hosts workspaces in Lunary's environment. Self-hosting keeps the monitoring deployment in your environment; it doesn't determine where the models your application calls run.

PII masking helps limit personal information in captured interactions. Role-based access controls and SSO govern access to monitoring data. Lunary is SOC 2 Type II and ISO 27001 certified.