lmnr-ai/lmnr

What it solves

Laminar is an observability platform designed to help developers monitor, debug, and evaluate AI agents. It addresses the complexity of tracking agent behavior, identifying failures, and managing datasets in production environments where traditional logging tools often fall short.

How it works

Laminar uses an OpenTelemetry-native SDK to automatically trace interactions across popular frameworks like LangChain, Vercel AI SDK, and various LLM providers. It captures agent execution data, which can be analyzed through custom dashboards, SQL queries, or automated signals that alert you when specific behaviors occur. The platform also provides tools for data annotation and running evaluations locally or in CI/CD pipelines.

Who it’s for

This platform is for developers and AI engineers building production-grade AI agents who need deep visibility into agent performance, trace debugging, and automated quality assurance.

Highlights

• Comprehensive Tracing: Automatically instruments major AI frameworks and LLM providers with minimal code.
• Agent Signals: Allows you to define and track specific agent behaviors (e.g., "stuck in a loop") with real-time Slack notifications.
• Evaluation Suite: Includes an extensible SDK and UI for running and comparing model evaluations.
• High Performance: Built in Rust with a custom engine that provides 20x trace compression for efficient storage and real-time visualization.
• Developer-Friendly Access: Supports SQL-based querying of traces and metrics, including an MCP/CLI interface for coding agents to debug issues autonomously.