Lunary

What is Lunary?

Lunary (also known as LLMonitor) is an observability and logging platform for AI agents and chatbots built on the LLM framework. It gives developers visibility into how their agents behave, how they perform, and how users interact with them.

Core features include analytics, tracing, request monitoring, cost tracking, conversation replay, debugging of agent executions, user activity tracking, training dataset creation, user feedback collection, and assertions to verify agent behavior. It integrates via an SDK and can be self-hosted or used as a hosted service.

Who created Lunary?

Lunary is the company behind LLMonitor.

What is Lunary used for?

• Monitor agent behavior and performance: Track requests and see what’s happening under the hood of AI applications.
• Track and manage costs: Understand costs per user and per model to optimize prompts and reduce expenses.
• Debug agent executions: Replay conversations and debug runs to pinpoint issues and see exactly what happened.
• Understand users: Track user activity and costs to identify the most active users and analyze behavior patterns.
• Build training datasets: Label outputs based on tags and user feedback to improve model quality.
• Collect feedback and verify functionality: Capture user feedback and run assertions to confirm agents work as expected.
• Flexible integration and hosting: Incorporate the platform via the SDK, then choose self-hosting or the hosted version.

Key takeaways

Lunary provides observability and analytics designed specifically for AI agents and chatbots. Developers use it to fine-tune applications, control costs, debug and replay agent runs, build better training data, and improve the user experience — with SDK-based integration and the choice of self-hosted or hosted deployment.