How developers build reliable AI agents.

The single open-source platform to trace, evaluate, and analyze AI agents.

Get started - free Read the docs

Backed by

ObservabilityEvalsPlaygroundLabeling

With Laminar, teams monitor agents in production, understand failure modes, and create evals to improve agent performance

Why teams choose Laminar

Fully Open-Source

Self-host or use our cloud. Transparent, extensible, and community-driven.

Highly scalable

Rust-powered, optimized for performance and scalability, capable of ingesting hundreds of millions of traces per day.

SQL access to all data

Analyze traces, metrics, and events with a built-in SQL editor. Bulk create datasets from queries.

Powerful SDK

With Laminar best-in-class SDK you can easily trace your agents and write powerful evals.

Custom dashboards

Turn SQL queries into custom dashboards to track custom metrics of your agent.

Affordable at scale

Start with a generous free tier, and scale without breaking the bank. No limit on the amount of spans you can ingest.

"Laminar's evals help us maintain high accuracy while moving fast. We now use them for every LLM based feature we build."

Hashim Rehman

CTO, Remo

OpenTelemetry LangGraph CrewAI\
\
CrewAI AI SDK 🚅\
\
LiteLLM Browser Use 🤘\
\
StageHand Playwright OpenAI icon\
\
OpenAI Anthropic Gemini\
\
Gemini Mistral\
\
Mistral Bedrock\
\
Bedrock Groq\
\
Groq

Integrate in 2 minutes

Simply initialize Laminar at the top of your project and popular LLM frameworks and SDKs will be automatically traced. Use our SDK to add comprehensive tracing to your agent.

OpenTelemetry LangGraph CrewAI\
\
CrewAI AI SDK 🚅\
\
LiteLLM Browser Use 🤘\
\
StageHand Playwright OpenAI icon\
\
OpenAI Anthropic Gemini\
\
Gemini Mistral\
\
Mistral Bedrock\
\
Bedrock Groq\
\
Groq

Observe & Debug

See traces of long-running agents in real time

Don't wait until the end of your AI agent run to start debugging. Laminar shows spans as they happen in real time.

Automatic error capture

Laminar automatically captures application level exceptions.

Tool calls and structured output tracing

Understand when your agent fails to use tools and to produce structured output.

Track custom metrics with events

You can emit events with custom metadata from your code to track custom metrics of your agent.

See what your browser agent sees

media loading

Error NaN

PlayPause

0:00

0:00

PlayPause10

Seek backward

10
10

Seek forward

10
0:01

MuteUnmute

Quality
1x

Playback rate

Audio

Captions

start airplaystop airplay

Start castingStop casting

Enter picture in picture modeExit picture in picture mode

Enter fullscreen modeExit fullscreen mode

Laminar automatically captures browser window recordings and syncs them with agent traces to help you see what the browser agent sees. Laminar automatically traces Browser Use, Stagehand and Playwright.

🤘

Learn about browser agent observability

Query & Analyze

Query all data on the platform with SQL

Access to traces, evals, datasets, and events data on the platform with a built-in SQL editor.

Track what matters to you with custom dashboards

Skip the dashboard builder. Just write SQL to create custom dashboards to track custom metrics of your agent.

From query to eval datasets in seconds

Use SQL to query custom data filtered by your own criteria. Batch insert to labeling queues or directly to datasets in seconds.

Access platform data via SQL API

Use the Laminar SQL API to query traces, evals, datasets, and events data from your own applications.

Evaluate & Iterate

Zero boilerplate evaluation SDK

Skip the setup hell. Write your agent function and evaluator, pass in your data, and run. We automatically handle parallelism and retries.

Iterate on prompts without touching your codebase

Open LLM calls in the Playground. Iterate fast - test new prompts, try different models, and validate improvements.

Catch regressions before your users do

See the impact of every change before it goes live. Compare evaluation runs to catch regressions early and validate that your improvements actually work.

Build high-quality eval datasets efficiently

No complex labeling tools or workflows. Just queue your data and start labeling. Perfect for teams getting started with systematic evaluation.

Start building reliable AI agents today.

Get started - free Read the docs

Contact us GitHub Join Discord Privacy policy Terms of service Status

Laminar