How do I know if LangSmith is down?

Check LangSmith status by: 1) Visiting the official LangChain status page for LangSmith incidents, 2) Testing a request against api.smith.langchain.com with your API key, 3) Checking whether traces are queuing locally in your SDK but failing to appear in the dashboard, or 4) Using API Status Check for real-time endpoint monitoring.

What should I do when LangSmith is down?

When LangSmith is unresponsive, your application traffic should keep working since most SDKs buffer traces asynchronously and drop them silently on failure rather than blocking your LLM calls. Check the status page, verify your API key, and temporarily disable tracing if buffering is causing memory pressure in high-throughput services.

How long do LangSmith outages usually last?

Trace ingestion delays typically clear within 10–20 minutes as backpressure resolves. Dashboard or evaluation outages tied to broader infrastructure incidents can take longer, though core LLM calls in your application are unaffected since tracing is decoupled from execution.

What are the best alternatives to LangSmith if it's down?

The best LangSmith alternatives for LLM observability include Langfuse (open-source, self-hostable), Helicone (simple proxy-based logging), Arize Phoenix (open-source tracing and evals), and Weights & Biases Weave (broader ML experiment tracking with LLM support).

Is LangSmith Down? How to Check LangChain LangSmith Status in 2026

Q: Why does LangSmith go down?

LangSmith disruptions are typically caused by trace ingestion backpressure during high-volume logging spikes, evaluation run queue backlogs, dataset storage issues, or underlying cloud infrastructure incidents affecting the hosted SaaS platform.

LangSmith is LangChain's hosted observability platform for tracing, debugging, and evaluating LLM applications — used widely by teams shipping agents and RAG pipelines built on the LangChain and LangGraph frameworks. Unlike a core inference API, LangSmith sits alongside your application rather than in the critical execution path, which changes what a "LangSmith is down" incident actually means for your users.

Whether you're seeing missing traces in the dashboard, stalled evaluation runs, or errors submitting feedback scores, this guide answers the key question: is LangSmith down for everyone, and is your production LLM traffic actually affected?

How to Check if LangSmith is Down (Fastest Methods)

1. Check the Official LangChain Status Page

LangChain publishes a status page covering LangSmith tracing, the evaluation service, and the dashboard UI. Check it first for any active incidents before digging into your own integration.

2. Test the LangSmith API Directly

A quick API test isolates whether the issue is on LangSmith's end or your integration:

curl https://api.smith.langchain.com/info \
  -H "x-api-key: $LANGSMITH_API_KEY" \
  -o /dev/null -w "%{http_code}"

A 200 response means LangSmith is reachable. A 401 means your API key is invalid. A 5xx or connection timeout means LangSmith is experiencing an outage.

3. Confirm Your Application Isn't Actually Broken

This is the most important step for LangSmith specifically: the official Python and JS SDKs submit traces asynchronously in the background and are designed to fail silently rather than block or crash your LLM calls. If LangSmith is down, your chatbot or agent should keep responding to users — only the tracing and evaluation layer is affected.

4. Check Local SDK Buffering Warnings

Review your application logs for LangSmith client warnings about failed submissions or growing local trace buffers. This confirms the tracing pipeline is degraded without you needing to speculate about the cause.

📡

Recommended

Monitor Your LLM Observability Stack

Don't fly blind on your LLM app's behavior. Get instant alerts when LangSmith trace ingestion stops working.

Try Better Stack Free →

Why Does LangSmith Go Down?

LangSmith's role as a high-volume telemetry sink for LLM applications creates its own set of failure modes:

Trace Ingestion Backpressure: Teams running high-throughput agents can generate large volumes of nested trace spans. Sudden spikes in trace volume can overwhelm ingestion capacity faster than autoscaling responds.
Evaluation Run Queue Backlogs: Running large-scale evaluation datasets against multiple models queues significant compute. Backlogs during peak usage can delay eval results without affecting live tracing.
Dataset Storage Issues: Datasets, example sets, and annotation queues are stored separately from live traces. Storage-layer incidents can affect dataset management while trace ingestion stays healthy, or vice versa.
Dashboard-Only Degradation: The web UI and the ingestion API are separate services. It's possible for traces to be ingesting normally while the dashboard itself is slow or erroring.
Underlying Cloud Provider Issues: As a hosted SaaS platform, LangSmith runs on cloud infrastructure that can experience regional incidents affecting latency or availability.

LangSmith Troubleshooting Checklist

Step 1: Identify the Error Type

HTTP 401 — API key invalid or expired. Regenerate in your LangSmith settings.
HTTP 429 — Rate limit exceeded on trace ingestion. Check your plan's trace volume limits.
HTTP 5xx / timeout — Platform-level outage. Check the LangChain status page.
Traces missing but no SDK errors — Possible dashboard-only degradation; ingestion may still be processing with delay.
Evaluation runs stuck "pending" — Compute queue backlog, typically resolves without action.

Step 2: Confirm Production Traffic Is Unaffected

Since tracing is decoupled from execution in the official SDKs, verify your chatbot, agent, or RAG pipeline is still serving real responses before treating a LangSmith outage as user-facing.

Step 3: Disable Tracing Temporarily if Needed

If a growing local trace buffer is causing memory pressure in a high-throughput service, set LANGCHAIN_TRACING_V2=false temporarily until LangSmith recovers, then re-enable it.

Step 4: Set Up Independent Monitoring

Don't rely solely on LangSmith's own status page — monitor the API endpoint directly so you know about ingestion failures even if the status page hasn't updated yet.

Best LangSmith Alternatives When It's Down

Langfuse

Open-source, self-hostable LLM observability with tracing, evals, and prompt management.

Helicone

Lightweight proxy-based logging for LLM calls, minimal integration overhead.

Arize Phoenix

Open-source tracing and evaluation toolkit with strong OpenTelemetry support.

Weights & Biases Weave

LLM tracing built on W&B's established experiment tracking infrastructure.

Conclusion: Tracing Should Never Take Down Your LLM App

LangSmith gives LangChain teams visibility they can't get otherwise, but it's built to fail safely — a LangSmith outage should degrade your observability, not your product. Teams that verify this decoupling and monitor the ingestion endpoint independently spot real incidents faster and avoid false alarms.

Get LangSmith Outage Alerts Instantly

Monitor LangSmith and your entire LLM observability stack. Get instant Slack or email alerts the moment trace ingestion fails.

Start Your Free Trial →