An open-source observability platform purpose-built for AI agents, providing tracing, monitoring, evaluation, and SQL analysis capabilities.
Laminar is an open-source observability platform purpose-built for AI agents, developed by a Y Combinator S24 alumni team backed by a $3M seed round. The platform is structured around four core modules for the agent lifecycle: the Tracing module, based on OpenTelemetry, enables one-line integration and features a unique True Agent Debugger that allows step-by-step replay with context retention, alongside synchronized screen recording for browser agents; the Signals monitoring module allows users to define event rules using natural language, automatically clustering anomalies and generating insights, with automated report delivery via email and Slack; the Evals module provides an agnostic SDK and CLI for custom metric regression testing locally or in CI/CD pipelines, with built-in data annotation and custom dataset creation; and the built-in SQL engine enables flexible querying and dashboard creation across all traces, metrics, and events data, with a standard SQL API for external systems.
Architecturally, Laminar adopts a decoupled front-end/back-end design with a Rust core engine. The data pipeline uses gRPC for high-efficiency transmission, relies on ClickHouse for underlying data persistence, and leverages a custom Rust query engine and real-time engine to ensure fast full-text search and streaming display of massive span data. Governed by the Apache License 2.0, the project supports lightweight or production-grade self-hosting via Docker Compose, alongside a managed cloud service. On the compliance front, the platform is HIPAA compliant and SOC2 Type 2 certification is currently in observation. It natively integrates with major development frameworks including Vercel AI SDK, LangChain, OpenAI, Anthropic, Gemini, Browser Use, Stagehand, and Playwright.
Quick Start
Self-hosted deployment:
git clone https://github.com/lmnr-ai/lmnr
cd lmnr
docker compose up -d
UI accessible at http://localhost:5667. For production, use docker-compose-full.yml.
TypeScript SDK:
npm add @lmnr-ai/lmnr
import { Laminar } from '@lmnr-ai/lmnr';
Laminar.initialize({ projectApiKey: process.env.LMNR_PROJECT_API_KEY });
Python SDK:
pip install --upgrade 'lmnr[all]'
from lmnr import Laminar
Laminar.initialize(project_api_key="<LMNR_PROJECT_API_KEY>")
Key Configuration
LMNR_PROJECT_API_KEY: Required project API key for SDK initialization.GOOGLE_GENERATIVE_AI_API_KEY: Required for Signals/AI monitoring (must be set on both app-server and frontend).baseUrl: Must be configured in the SDK when self-hosting to point to the self-hosted address and port.