laminar

Laminar raises $3M seed to build observability for long-running agentsRead moreBacked by Y CombinatorUnderstand why your agent failed.Iterate fast to fix it.Open-source observability platform for long-running agents.Trace in one line, debug from any step, detect patterns at scale.Get StartedRead the DocsTracingSignalsDebuggerEvalsSQLFeatures forevery step ofagent development01.TracingStart tracing your agent in minutesTwo lines to integrate with the AI frameworks or SDKs you useimport { openai } from '@ai-sdk/openai';import { generateText } from 'ai';import { getTracer } from '@lmnr-ai/lmnr';const { text } = await generateText({ model: openai('gpt-4.1-nano'), prompt: 'What is Laminar flow?', experimental_telemetry: { isEnabled: true, tracer: getTracer(), },});DocsTrue Agent DebuggerFirst-of-its-kind agent developer experienceRun local, debug in browserRerun at step N with previous context preservedTune your system promptsInstantly reflect changes as you saveDocsagent.runnavigatenavigated to https://laminar.shagent.stepgemini-2.5-flashI can see the Laminar homepage. Let me click on the Pricing link in the navigation to view their pricing plans.clickclicked 'Pricing' linkagent.stepgemini-2.5-flashThe pricing page has loaded. I can see three tiers: Free, Hobby, and Pro. Let me extract the details.extractextracted pricing tier informationagent.stepgemini-2.5-flashExtracting pricing information. The Free tier offers 1GB data with 15-day retention.extractextracted Free tier detailsagent.stepgemini-2.5-flashContinuing extraction. The Hobby tier is $25/month with 2GB data and 30-day retention.extractextracted Hobby tier detailsagent.stepgemini-2.5-flashFinal extraction complete. I've gathered all pricing data and will now compile the summary.write_filewriting to pricingsummary.mdagent.stepgemini-2.5-flashNow I need to navigate to the docs page to gather additional information about the API.clickclicked 'Docs' linkagent.stepgemini-2.5-flashThe documentation page is now loaded. I can see sections for Tracing, Evaluations, and SDK reference.extractextracted documentation structureagent.stepgemini-2.5-flashLet me explore the SDK reference to understand the available methods and integrations.clickclicked 'SDK Reference'agent.stepgemini-2.5-flashFound the Python SDK documentation. It shows initialize(), observe(), and evaluate() as main methods.extractextracted SDK method signaturesagent.stepgemini-2.5-flashI have gathered all necessary information. Writing the final summary report with pricing and SDK details.write_filewriting to final_report.md0s10s20s30s40s50s60s70s80s90s100sFull trace context at a glanceGet full context of what your agent was doing and where it went wrong without digging through hundreds of spans.DocsExplain the error that's happeningDiagnose the core cause of the issueSummarize my trace for me and explain the root cause of the errorSummarize my trace for me and explain the root cause of the errorConvoluted trace? Ask AIAsk our AI agent to summarize, analyze, and debug your trace no matter the complexity.DocsSession replayfor browser agentsLaminar captures browser screen recordings and automatically syncs them with agent traces. Easily integrates with Browser Use, Stagehand, Playwright, Kernel, Browserbase, and more.Docs02.AnalysisSignals answer any question, from any trace, at scaleDescribe a Signal you're looking for, Laminar extracts it from past and future traces.Analyze millions of tracesdown to the individual spanDefine what you're looking for andspecify the output formatMy agent failureDefinitionOutput SchemaNameTypeDescriptionCategoryEnumCategory of the failureDetailsStringDetails of what failed and whyToolsNumberNumber of tools usedSignals agent detects events fromtraces based on your definitionEventsTimestampCategoryDescriptionJan 9, 8:47tool_errorThe tool get_shared_data failed to return valid results for the query...Jan 9, 11:00api_errorThe agent attempted to fetch via curl on the remote endpoint but received...Jan 9, 14:13logic_errorAn incoming user request was misidentified as a system command due to...Jan 9, 17:26timeoutThe second tool call made on the database mcp server timed out after...Jan 9, 20:39validation_errorInput validation failed on the request body parameter for the batch...Jan 10, 23:52tool_errorThe tool parse_json returned malformed output when processing nested...Jan 10, 2:05api_errorAPI rate limit exceeded while attempting batch processing of embeddings...Jan 10, 5:18logic_errorThe routing logic incorrectly classified the intent as a search query...Jan 10, 8:31timeoutConnection timeout when reaching the external auth provider endpoint...Jan 10, 11:44validation_errorSchema validation rejected the response format from the completions...Jan 11, 14:57tool_errorThe tool write_file encountered permission denied on the output directory...Jan 11, 17:10api_errorREST API returned 503 during the health check sequence for service...Jan 11, 20:23logic_errorConditional branch took wrong path due to null comparison in filter...Jan 11, 23:36timeoutRequest to embeddings service exceeded 30s timeout during peak load...Jan 11, 2:49validation_errorField type mismatch in the aggregation pipeline stage caused silent...Jan 12, 5:02tool_errorThe tool search_index failed with corrupted query vector from cache...Jan 12, 8:15api_errorOAuth token refresh endpoint returned invalid grant error on rotation...Jan 12, 11:28logic_errorLoop detection triggered false positive on recursive tool invocation...Jan 12, 14:41timeoutDatabase connection pool exhausted during peak load causing cascade...Jan 12, 17:54validation_errorParameter bounds check failed for the pagination offset in list query...Jan 13, 20:07tool_errorThe tool get_shared_data failed to return valid results for the query...Jan 13, 23:20api_errorThe agent attempted to fetch via curl on the remote endpoint but received...Jan 13, 2:33logic_errorAn incoming user request was misidentified as a system command due to...Jan 13, 5:46timeoutThe second tool call made on the database mcp server timed out after...Jan 13, 8:59validation_errorInput validation failed on the request body parameter for the batch...Jan 14, 11:12tool_errorThe tool parse_json returned malformed output when processing nested...Jan 14, 14:25api_errorAPI rate limit exceeded while attempting batch processing of embeddings...Jan 14, 17:38logic_errorThe routing logic incorrectly classified the intent as a search query...Jan 14, 20:51timeoutConnection timeout when reaching the external auth provider endpoint...Jan 14, 23:04validation_errorSchema validation rejected the response format from the completions...All events are clusteredfor high-level insightsClusters1Incorrect tool call02Confused user03API call failed0Jan 12Jan 13Jan 14Jan 15Receive insights about your tracesautomatically in email and Slackfrom: reports@mail.lmnr.aiSignals ReportTotal events365AI Startup · Signals Events SummaryMar 06, 2026 - Mar 13, 2026My AgentSignalEventsBrowser errors11Slow tasks and optimization246Server startup errors94SummaryThe agent frequently performs inefficient sequential edits on dozens of files instead of using scripted automation, leading to high latency and redundant tool calls. Sandbox environments often require manual Git authentication and dependency setup.Laminar3:18 pmEvent:Agent failureCategory:logic_errorDescription:The LLM in the 'refine_report' task failed to follow the instruction to keep the summary to 3-4 sentences.DocsPlatform-wide SQLQuery all platform data with SQLFeed your Evals with Datasets straight from SQL queriesSQL API to pull data into your applicationDocsSQL Editor1234SELECT *FROM spansWHERE trace_id = 'e44f93ea-35f5-d9d8-1dc1-ae29863504a9'ORDER BY start_time ASCRunspan_idnamespan_typestart_timeend_timeduration00000000-0000-00...agent.runDEFAULT2025-12-16 21:48:18.502025-12-16 21:48:18.50May 05, 5:3000000000-0000-00...agent.runDEFAULT2025-12-16 21:48:18.502025-12-16 21:48:18.50May 05, 5:3000000000-0000-00...agent.runDEFAULT2025-12-16 21:48:18.502025-12-16 21:48:18.50May 05, 5:3000000000-0000-00...agent.runDEFAULT2025-12-16 21:48:18.502025-12-16 21:48:18.50May 05, 5:3000000000-0000-00...agent.runDEFAULT2025-12-16 21:48:18.502025-12-16 21:48:18.50May 05, 5:3000000000-0000-00...agent.runDEFAULT2025-12-16 21:48:18.502025-12-16 21:48:18.50May 05, 5:3000000000-0000-00...agent.runDEFAULT2025-12-16 21:48:18.502025-12-16 21:48:18.50May 05, 5:3000000000-0000-00...agent.runDEFAULT2025-12-16 21:48:18.502025-12-16 21:48:18.50May 05, 5:30Custom dashboardsTrack tokens, latency, and more.Correlate data across users and sessions.Comprehensive UI dashboard builder with support for custom SQL queries.Docs03.EvalsRobust EvalsVerify progress, catch regressions, and iterate with confidence123456789101112131415evaluate({ data: dataset, executor: capitalOfCountry, evaluators: { accuracy: (output: string, target: string | undefined): number => { if (!target) return 0; return output.includes(target) ? 1 : 0; }, }, config: { instrumentModules: { openAI: OpenAI } }})Evals SDK you want to useDefine your agent, dataset, and success metric. We handle the rest.RewardAverage0.72Trace123.36s81k0.005ReaderFiltersSearchMetadataAsk AInavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows we are on https://laminar.sh with several top navigation links, including a visible Pricing link at index 4. The user asked to go to laminar.sh and describe their pricing plans. The immediate next actionable step is to navigate to the Pricing page to access the pricing information. I will click the Pricing link (index 4) to load the pricing page, after which I will extract and summarize the pricing plans on the page in the next step.clickgpt-05-nano-2025-08-07Current task: on the Pricing page for laminar.sh and need to extract and describe pricing plans. The previous step shows the Pricing link was clicked and the page likely loaded. Based on browser_state, there are multiple pricing tiers listed: Free, Hobby, Pro, Enterprise, plus a Pricing calculator with Free tier and token-based pricing. I will perform a structured extraction of the pricing sections to summarize plan names, prices, included features, data retention, and team members. This will prepare a clear description for the user. No navigation changes needed; just extract the page content relevant to pricing for accurate description.extracted pricing plans and their details on Laminar Pr...gpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysTeam members: 1Support: Community supportHobbyPrice: $30 / monthData: 3 GB data / month includedAdditional data: $2 per 1 GBData retention: 30 daysTeam members: UnlimitedSupport: Priority email supportProPrice: $150 / monthgpt-05-nano-2025-08-07Reviewing the agent history shows step-by-step progress: the user asked to go to laminar.sh and describe pricing plans. We navigated to the pricing page and extracted structured details about Free, Hobby, Pro, Enterprise, and the pricing calculator. The read_state_0 provides a concise, structured summary including price, data, retention, team members, and support for each plan. Next, I will consolidate this into a readable summary file for easy reference and provide a concise description to the user. Since the instruction prefers stepwise actions, I'll save the structured summary to a file for traceability and then prepare the user-facing description in the next step.writing to pricingsummary.mdAs granular as you wantSee high level results or dive deep into your individual traces.Docs“We run millions of agent sessions in our cloud, and when something goes wrong, Laminar's trace view is the first place we look. ”Magnus MüllerCEO, Browser UseSelf-host anywhereFully open-source, Apache 2.0 licensedSet up with Docker in three linesDeploy anywhere with Helm chartsDocsProduction-gradeto the coreWritten in Rust for blazing fast performanceTerabytes of data with easeHIPAA compliant, SOC2 Type 2 (in observation)ComplianceTrace123.36s81k0.005ReaderFiltersSearchMetadataAsk AInavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysnavigated to https://laminar.shgpt-05-nano-2025-08-07Current state shows navigation links including Pricing.clickgpt-05-nano-2025-08-07Extracting pricing information from the page...extracted pricing plansgpt-05-nano-2025-08-07FreePrice: $0 / monthData: 1 GB data / monthData retention: 15 daysgit clone https://github.com/lmnr-ai/lmnrcd lmnrdocker compose up -dUnderstand why your agent failed.Iterate fast to fix it.Read the DocsGet StartedContact UsGitHubJoin DiscordPrivacy PolicyTerms of ServiceStatus --- Laminar raises $3M seed to build observability for long-running agentsRead moreFree$0 / month1 GB datano overage100 signal runsno overage15 day retention1 project1 seatCommunity supportGET STARTEDHobby$30 / month3 GB data includedthen $2 / GB1,000 signal runs includedthen $0.02 / run30 day retentionUnlimited projectsUnlimited seatsEmail supportGET STARTEDPro$150 / month10 GB data includedthen $1.50 / GB10,000 signal runs includedthen $0.015 / run90 day retentionUnlimited projectsUnlimited seatsSlack supportGET STARTEDEnterpriseCustomCustom limitsOn-premiseUnlimited projectsUnlimited seatsDedicated supportCONTACT USPricing calculatorTokens per month100M≈ 0.3 GBSignal runs per month100FreeRecommendedBase$0.001 GB + 100 runs includedData (0.3 GB)IncludedSignals (100)IncludedTotal$0.00/mo15-day retentionCommunity supportFrequently Asked QuestionsHow is data usage calculated?What is a signal run?Understand why your agent failed.Iterate fast to fix it.Read the DocsGet StartedContact UsGitHubJoin DiscordPrivacy PolicyTerms of ServiceStatus --- Laminar raises $3M seed to build observability for long-running agentsRead moreBlogWe raised $3M to build observability for long-running agentsMar 16, 2026·Robert Kim Your agent ran for 40 minutes and failed. You spend 20 minutes digging through logs, but still have no clue what went wrong. Now multiply that by a thousand runs in production, and you have a second problem: you don't even know which runs are failing, or what patterns are emerging across them. We built Laminar to fix this. Today we're announcing our $3M seed round led by Atlantic.vc, with participation from Y Combinator, AAL.vc, and angels including Ben Sigelman (co-creator of OpenTelemetry) and Ant Wilson (CTO of Supabase). Where existing tools fall short# Most observability platforms are, at their core, data collectors. They ingest traces and metrics, store them, and then leave it entirely up to you to figure out what's going on. You get a tree of spans. You click into one, read it, go back, click the next one, try to hold the context in your head. For an agent that ran for 30 minutes and made 200 decisions, this simply doesn't work. What Laminar does# Instead of just collecting data and leaving you to make sense of it, Laminar actively helps you understand what happened in a single run, reproduce and fix it, and detect what's going wrong across thousands of runs in production. Tracing that actually helps you understand what happened.# With a single line of code, Laminar SDK auto-instruments the vast majority of AI frameworks and SDKs, including Claude Agent SDK, AI SDK, LiteLLM, Browser Use, Stagehand, OpenHands SDK, and more. Laminar's tracing UI gives you as much information as quickly as possible. Our trace timeline and reader mode lay out the agent's reasoning and actions as a clean, readable feed, instead of tree of spans. For browser agents, we record full browser sessions synced with traces so you can see exactly what the agent saw at every step. If a trace is too complex to parse visually, you can chat with it in natural language instead of manually digging through hundreds of steps. A debugger that doesn't make you start from scratch.# When your agent fails 15 minutes into a run, the normal workflow is: restart from scratch, wait for it to reach the same state, hope it reproduces the failure. Laminar's debugger lets you go to the exact step where things went wrong, tweak your prompt or tool definitions, and rerun from that step with full context preserved. Under the hood, when you rerun from a step, we mock all prior LLM calls by replaying their original responses. The agent walks through its prior steps instantly without spending any tokens, and external state (browser DOM, sandbox environment, conversation history) is fully restored. By the time execution reaches your breakpoint, everything is exactly as it was. Signals# In production you have thousands of runs and the question is no longer "why did this run fail" but "what's going wrong across all my runs." With Signals, you can describe what you want to detect in natural language ("agent gets stuck in a retry loop," "user rephrases out of frustration") and Laminar extracts matching events from every trace, then clusters them into patterns. All of your data is accessible through a built-in SQL editor, both in the UI and via API. Run arbitrary queries against your traces, spans, and events. Build custom dashboards, do ad-hoc analysis, or bulk-create datasets from production traces. Those datasets plug directly into our evals pipeline, so you can run evaluations on real production data. Who's using it# Teams building some of the most complex agents in production rely on Laminar today, including Browser Use, OpenHands, Rye.com, Alai, LegionIntel, and many more. Try it# Laminar is fully open source and extremely fast (written in Rust). Self-host it anywhere: git clone https://github.com/lmnr-ai/lmnr cd lmnr docker compose up -d Or sign up on our managed platform and start tracing in under a minute.Understand why your agent failed.Iterate fast to fix it.Read the DocsGet StartedContact UsGitHubJoin DiscordPrivacy PolicyTerms of ServiceStatus --- Laminar raises $3M seed to build observability for long-running agentsRead moreWe raised $3M to build observability for long-running agentsMar 16, 2026Migrate from Helicone to LaminarMar 5, 2026Migrate from Traceloop to LaminarMar 5, 2026Laminar vs BraintrustFeb 17, 2026Laminar vs LangfuseFeb 13, 2026Rebuilding My First Startup with Claude Agent SDKFeb 10, 2026The Anatomy of a Wasted Credit: Why AI App Builders Need Trajectory AnalysisFeb 5, 2026Laminar vs Langfuse vs LangSmith: LLM Observability Compared (2026)Jan 29, 2026Introducing Signals — get insights from agent traces at scaleJan 26, 2026To Scaffold or Not to Scaffold: The Problems That Won't DissolveJan 26, 2026Debugging Parallel AgentsJan 20, 2026You Can Now Trace Stagehand Workflows and Agents End-to-EndDec 8, 2025How we instrumented Claude Agent SDK using a tiny Rust proxyDec 3, 2025Understand why your agent failed.Iterate fast to fix it.Read the DocsGet StartedContact UsGitHubJoin DiscordPrivacy PolicyTerms of ServiceStatus