Laminar
Open-source, OpenTelemetry-native observability platform for AI agents.
About
Laminar (lmnr-ai/lmnr) is an open-source observability platform built specifically for AI agents, tracing every agent run, model turn, tool call, and subagent spawn with a single Laminar.initialize() call. It is OpenTelemetry-native, with a TypeScript + Rust real-time engine, and ships deep integrations for LangChain deepagents, Pydantic AI, Mastra, the Claude Agent SDK, Temporal workflows, and a one-line bridge that dual-exports Langfuse spans. The repo has roughly 3,000 GitHub stars and the company is a Y Combinator S24 startup founded by Robert Kim.
In March 2026 Laminar raised a $3 million seed round led by Atlantic with participation from Y Combinator and AAL VC, and it is already used by organizations running large-scale agent deployments such as Browser Use and OpenHands. Teams can self-host via Helm or use the managed cloud.
Limitation: the agent-observability space is crowded (Langfuse, LangSmith, Braintrust, Arize all overlap), so teams already instrumented on another platform should weigh the switching cost; the Langfuse bridge helps, but it is Python-only.
What it can do
- OpenTelemetry-native tracing of agent runs, tool calls, and subagents
- One-line integrations for LangChain, Pydantic AI, Mastra, Claude Agent SDK
- Langfuse dual-export bridge for side-by-side evaluation
- Self-hostable via Helm plus a managed cloud tier
Setup
Self-host via Helm or sign up for Laminar Cloud; instrument with Laminar.initialize() in Python or TypeScript.
Prerequisites
An AI agent or LLM pipeline to trace; Python or TypeScript runtime.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 61; Trending score 80/100 from researcher assessment.
★ 2,954 GitHub stars
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.