LangSmith

What it does

LangSmith Observability takes in execution data from an AI agent or LLM application and produces step-by-step traces, dashboards, alerts and automatic summaries of usage patterns. Traces arrive through native support for popular agent frameworks, OpenTelemetry, or SDKs for Python, TypeScript, Go and Java. Outputs include cost, latency, token and error views, quality scores from online evaluations, and clustered topics and failure modes.

Teams use it to find which step of a multi-step agent run is slow, expensive or wrong, then watch production quality over time. Monitoring supports LLM-as-judge and code evaluators, tool and agent trajectory monitoring, and webhook or PagerDuty alerts. It is framework-agnostic, so applications built with the OpenAI SDK, Anthropic SDK, Vercel AI SDK, LlamaIndex or custom code can be traced. SmithDB, its trace store, can be self-hosted inside your own VPC.

It suits developers and teams building agents who need visibility after deployment. Sign-up at smith.langchain.com is required; the Developer plan includes one seat and 5k base traces a month, and paid usage follows. Base traces are kept 14 days, and longer 180-day retention costs extra. Self-hosted and hybrid options need the Enterprise plan. Some features, such as Tuned Evaluators, are in public beta with plan and region limits.

What it can do

• Step-by-step agent tracing with message threading for multi-turn chats
• Dashboards for token usage, latency, error rates, cost and feedback scores
• Online LLM-as-judge and code evaluators for production traffic
• Webhook and PagerDuty alerts when metrics cross thresholds
• Automatic trace clustering and error-analysis insights
• OpenTelemetry ingest and export for existing observability pipelines

Works with

• OpenTelemetry
• OpenAI SDK
• Anthropic SDK
• Vercel AI SDK
• LlamaIndex
• PagerDuty

Example workflows

Find a slow or costly agent step

Input: an agent app instrumented with a LangSmith SDK or OpenTelemetry. Steps: send traces, open a trace to inspect each LLM and tool call, filter by latency or cost. Output: the specific run and step causing the problem.

Monitor production quality with alerts

Input: live traces from a deployed agent. Steps: build a dashboard of latency and cost, add an online LLM-as-judge evaluator, set a webhook or PagerDuty alert. Output: scored traces and notifications when metrics cross a threshold.

Before you choose

• Base traces are retained 14 days; extended 180-day retention costs more
• Self-hosted and hybrid deployment require the Enterprise plan
• Cloud data is stored in GCP us-central-1 on the hosted service; the pricing page also lists US or EU cloud

Listed under

• Coding
• Coding agents
• DevOps and observability
• Agent monitoring
• AI engineering
• AI teams
• Developers
• Evaluation
• LLM tracing

Get started

Developer plan; demo available

1. Create an account at smith.langchain.com
2. Install a Python, TypeScript, Go or Java SDK, or point OpenTelemetry at LangSmith
3. Run your application so traces appear, then build dashboards and alerts

Developer and Plus seats are billed monthly; traces are billed monthly in arrears. Enterprise is invoiced annually. Per-trace rates beyond included volume were not confirmed in the reviewed pricing information.

Source: langchain.com/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• LangSmith: Agent & LLM Observability Platform ↗www.langchain.com
• LangSmith Plans and Pricing ↗www.langchain.com
• Home - Docs by LangChain ↗docs.langchain.com

Company website

langchain.com: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.