Laminar

Laminar is an open-source observability platform designed to help developers monitor, debug, and improve the reliability of AI agents. It addresses the challenge of detecting, diagnosing, and resolving failures in agent-based systems by providing comprehensive visibility into agent runs and surfacing actionable insights when issues occur.

The platform offers real-time alerts when agents encounter failures, with a system called Signals that lets users describe errors in plain English. Laminar reads each agent run and notifies users, including via Slack, when specified issues arise. It provides a clear and navigable transcript and timeline of agent activity, highlighting inputs, LLM reasoning steps, tool calls, and interactions with sub-agents. Users can investigate issues by asking questions directly about the agent run, receiving answers linked to specific contextual points for rapid troubleshooting.

Laminar groups related events into named clusters, tracks their recurrence over time, and automatically resolves clusters when issues stop appearing, reopening them if problems return. It includes a CLI and MCP integration, enabling coding agents to autonomously iterate: running agents, reading traces, applying fixes, and rerunning with cached state. The Debugger session UI allows users to review these iterations in detail.

The platform supports the creation of evaluation datasets from resolved error clusters, enabling automated regression testing after code changes. Laminar offers integrations with a wide range of agent development and AI frameworks, including Vercel AI SDK, Claude Agent SDK, OpenAI Agents SDK, LangChain, DeepAgents, OpenCode SDK, LiteLLM, Mastra, Pydantic AI, OpenHands SDK, Stagehand, Playwright, OpenAI SDK, Gemini API, and Anthropic SDK. It provides features such as rerunning agents from any step, building custom dashboards with SQL queries, full SQL access to platform data, fast data annotation tools, screen recording for browser agents, and full-text search across all agent run data.

Laminar is available as an open-source solution and offers a free tier to help teams quickly integrate observability into their AI agent development workflows.