Laminar

Open-source, OpenTelemetry-native observability and evals built for AI agents

Hand-reviewed by the Toolglade team·Updated Aug 2026

llm-observability#llm-observability#agent-tracing#opentelemetry#evals

Free plan Claimed API Self-hosted Teams

Toolglade’s take

Laminar is a compelling, developer-first observability tool for teams building agents rather than simple chat apps, with genuinely useful touches like plain-English Signals and SQL access to all trace data. Being OpenTelemetry-native and Apache 2.0 open source makes it easy to adopt and self-host without lock-in. It is younger and smaller than the biggest observability players, so the ecosystem and integrations are still growing. Verify current pricing tiers and retention limits with the vendor.

About Laminar

Laminar is an open-source, OpenTelemetry-native observability platform for AI agents, offering one-line tracing, plain-English Signals, an evals framework, and SQL access to all trace data.

Laminar is an open-source observability and evaluation platform built specifically for AI agents and LLM applications. Backed by Y Combinator (S24) and written in Rust for performance, it provides an OpenTelemetry-native tracing SDK that instruments agents with essentially one line of code, automatically capturing traces from popular stacks like the Vercel AI SDK, LangChain, OpenAI, Anthropic, Gemini, and browser-automation tools such as Browser Use and Stagehand. Beyond tracing, Laminar adds features tailored to agentic workloads. Signals let developers describe any behavior they want to track in plain English, for example detecting when an agent is stuck in a loop, so teams can monitor qualitative agent behaviors rather than just raw spans. It also ships an unopinionated evals SDK and CLI for running evaluations locally or in CI/CD, with a UI to visualize and compare results, and it exposes all platform data, including traces, spans, signal events, and evaluations, through SQL for flexible analysis. Laminar uses transparent, data-volume-based pricing with no seat fees or per-span counting: a free tier of 1GB with short retention, a Hobby plan around $30/month, and a Pro plan around $150/month with longer retention. Because it is fully open source under Apache 2.0, teams can also self-host it in minutes via the official Helm chart. The company raised $3M to build observability for long-running agents.

TL;DR

Laminar is an open-source, OpenTelemetry-native observability and evals platform purpose-built for AI agents, with one-line tracing, plain-English Signals, and SQL access to all trace data.

Company overview

Laminar (lmnr) is a Y Combinator (S24) company building observability for AI agents, written in Rust for performance. It raised $3M to focus on observability for long-running agents.

The project is fully open source under Apache 2.0, reflecting a developer-first, no-lock-in philosophy, and positions itself as an alternative to established LLM observability tools with a specific emphasis on agentic workloads.

Product features

Laminar offers an OpenTelemetry-native tracing SDK that instruments apps with one line and auto-traces popular frameworks. Signals let developers track agent behaviors in plain English, and a built-in evals SDK and CLI run evaluations locally or in CI/CD with a comparison UI.

All platform data, including traces, spans, signal events, and evaluations, is queryable via SQL. Pricing is transparent and volume-based with no seat fees, and the platform can be self-hosted via Helm.

Target market

Laminar targets developers and teams building AI agents and LLM applications who need tracing, behavior monitoring, and evals without heavy per-seat costs or lock-in.