Laminar

Laminar is an open-source, OpenTelemetry-native observability and debugging platform built for AI agents. Trace, debug, and monitor every LLM call, tool call, and sub-agent your agent runs, whether it finishes in one turn or works through hundreds of steps across parallel sub-agents. Use managed Laminar Cloud or self-host the whole platform.

Agents produce enormous amounts of data, and a tree of span names tells you little about what the agent actually did. Laminar is built to turn that data into answers:

• During development, the debugger records a run and replays it with everything before your change served from cache, so each iteration takes seconds instead of a full live run. Your coding agent can drive the whole loop.

• In production, Signals let you describe outcomes and failures in plain language and extract structured events across all your traces, so you’re notified when your agent breaks and know why in seconds.

Tracing

Trace LLM calls, tool use, and custom functions. First-class support for AI SDK, Claude Agent SDK, LangChain, Browser Use, and more.

Viewing traces

Read each trace as a transcript: agent inputs, LLM turns, tool calls, and sub-agents collapsed into cards. Not a span tree.

Debugger

Record a run, replay it from cache up to the point you’re testing, and prove the fix with an eval. Built for a coding agent to run.

Signals

Describe outcomes and failures in plain language. Laminar reads every trace and produces structured events you can query, cluster, and alert on.

Evaluations

Run evals against datasets locally or in CI. Catch regressions before they ship.

Datasets & Labeling

Build evaluation datasets from production traces.

Playground

Replay any traced span. Swap prompts or models and compare results side by side.

Platform

Realtime traces, full-text search, raw SQL access from the UI, CLI, or MCP, and custom dashboards.