LangSmith

Free plan APILinuxMacSelf-hostedWebWindows

What LangSmith is

LangSmith is a platform for building, evaluating, deploying, monitoring, and improving AI agents. Teams can evaluate curated datasets offline or score production interactions in real time. Evaluation options include human review, heuristic checks, LLM-as-judge assessments, pairwise comparisons, and custom evaluators written in Python or TypeScript. Annotation queues let teams assign runs to subject-matter experts and use their feedback to tune evaluations, prompts, or datasets. The Playground supports prompt experiments and comparisons across versions or model providers, and LangSmith chat can improve prompts. For observability, it offers agent tracing, cost tracking, trajectory monitoring, evaluations, and webhook or PagerDuty alerts. It accepts traces through SDK or API from LangGraph, custom Python, or other stacks, and its SDKs include Python, TypeScript, Go, and Java. It supports pytest, Vitest, and GitHub workflows for evaluation pipelines. The Developer plan is free at $0 per seat per month; Plus is $39 per seat per month, then pay as you go. Enterprise pricing is custom.

Who it suits

LangSmith suits teams building or operating AI agents that need evaluation, tracing, prompt tools, or deployment choices. It offers cloud hosting for Developer and Plus, and cloud, hybrid, or self-hosted options for Enterprise.

What it does well

Offline and real-time evaluation options
Human, heuristic, LLM, pairwise, and custom evaluators
Tracing includes cost tracking and trajectory monitoring
Free Developer plan with up to 5k base traces monthly
Enterprise offers self-hosted and hybrid deployment

What to know first

Base traces have 14-day retention
Extended 180-day retention costs extra
Developer includes one seat and community support

Limits

Developer includes one seat and up to 5k base traces per month, after which usage is pay as you go. Plus includes up to 10k base traces per month, then pay as you go. Base traces are retained for 14 days; extended traces have 180-day retention and incur an additional fee to upgrade.

Plans

Plan	Billed and limits	Price
Developer	$0 per seat per month; usage pay-as-you-go · 1 seat · 5,000 base traces/month · community support	Free
Plus	$39 per seat per month; usage pay-as-you-go · 10,000 base traces/month · unlimited seats · 1 free small serverless deployment	$39 a month
Enterprise	invoiced annually upfront; usage pay-as-you-go · self-hosted or hybrid deployment · custom SSO, ABAC and RBAC · support SLA	On request

As langchain.com lists them, read 15 Oct 2026. The maker's own page has the final word.

Getting started

Open the LangSmith evaluation website.
Choose the Developer, Plus, or Enterprise plan.
Use the web platform or connect through the API or an SDK.
Send application traces through the SDK or API; LangSmith supports LangGraph and other stacks.
Set up evaluation workflows with supported tools such as pytest, Vitest, or GitHub workflows.

Verdict

LangSmith combines agent evaluation with prompt tools and observability, including cost and trajectory tracking. Teams should account for trace-retention limits and the pay-as-you-go structure when comparing plans.

Questions

Is LangSmith free?

Yes. The Developer plan costs $0 per seat per month, with usage charged after its included trace allowance.

How much does Plus cost?

Plus costs $39/user/mo, billed at $39 per seat per month, then pay as you go.

What platforms does LangSmith support?

The listed platforms are web, API, and self-hosted. Enterprise hosting options include cloud, hybrid, and self-hosted.

Which evaluation methods does it support?

It supports human review, heuristic checks, LLM-as-judge evaluators, pairwise comparisons, and custom evaluators in Python or TypeScript.

Which tools can connect to its evaluation workflows?

The evaluation page lists pytest, Vitest, and GitHub workflows for CI/CD evaluations.

How long are traces retained?

Base traces have 14-day retention. Extended traces have 180-day retention, with an additional fee to upgrade from base traces.

Facts from LangSmith's own pages

API and CLI: LangSmith exposes a REST API and a CLI that manages projects, traces, runs, datasets, evaluators, experiments, threads and sandboxes. · github.com, 15 Oct 2026
CI/CD: The evaluation workflow integrates with pytest, Vitest and GitHub workflows for automated regression checks. · langchain.com, 15 Oct 2026
CLI installation: The CLI provides installation paths for macOS, Linux and Windows through shell scripts, PowerShell, Homebrew, Scoop, Go or downloadable binaries. · github.com, 15 Oct 2026
Company origins: LangChain says the company began in early 2023 and that LangSmith launched first as a commercial observability product before expanding into evaluation and deployment. · langchain.com, 15 Oct 2026
Company profile: LinkedIn lists LangChain as privately held, headquartered in San Francisco, with 201–500 employees and 469 employees shown on the page. · linkedin.com, 15 Oct 2026
Data residency: Cloud data is hosted in GCP US or EU regions, while self-hosted data remains in the customer's VPC. · langchain.com, 15 Oct 2026
Data use: LangChain states that LangSmith data is not used to train models and remains owned by the customer. · langchain.com, 15 Oct 2026
Deployment: Deployment provides managed or self-hosted agent servers with streaming, cron scheduling, authentication, MCP support and more than 30 API endpoints. · langchain.com, 15 Oct 2026
Documentation: LangChain publishes reference documentation, integration guides, deployment instructions, a CLI guide and a REST API reference. · docs.langchain.com, 15 Oct 2026
Evaluation: Evaluation supports offline datasets, online production evaluations, human annotation queues and side-by-side experiment comparison. · langchain.com, 15 Oct 2026
Evaluators: Its evaluation framework supports human review, heuristic checks, LLM-as-judge evaluators, pairwise comparisons, and custom evaluators in Python or TypeScript. · langchain.com, 5 Oct 2026
Evaluator types: Evaluators can be human, heuristic, LLM-as-judge, pairwise or custom Python and TypeScript code. · langchain.com, 15 Oct 2026
Founded: 2023 · langchain.com, 28 Sept 2026
Framework support: LangSmith is framework-agnostic and accepts traces through its SDK or API from LangGraph, custom Python, or other application stacks. · langchain.com, 5 Oct 2026
Funding: LangChain announced a $125 million Series B at a $1.25 billion valuation on October 20, 2025. · langchain.com, 15 Oct 2026
Headquarters: San Francisco, California, United States · langchain.com, 28 Sept 2026
Hosting: The pricing page lists Cloud hosting for Developer and Plus, with Cloud, Hybrid, or Self-Hosted options for Enterprise. · langchain.com, 5 Oct 2026
Human feedback: Annotation queues let teams assign runs to subject-matter experts and use feedback to calibrate evaluations, improve prompts, or augment datasets. · langchain.com, 5 Oct 2026
Integrations: Documented integrations cover OpenAI, Anthropic, Bedrock, Gemini, Mistral, LiteLLM, OpenTelemetry, CrewAI, Vercel AI SDK, PydanticAI, AutoGen and many other frameworks and providers. · docs.langchain.com, 15 Oct 2026
Legal entity: Crunchbase identifies the maker as LangChain, Inc., a privately held San Francisco company founded by Harrison Chase and Ankush Gola. · crunchbase.com, 15 Oct 2026
Limit: deployment documentation: A LangChain forum user reported authentication failures and difficulty understanding deployment secrets, assistant IDs and setup documentation. · forum.langchain.com, 15 Oct 2026
Limit: large datasets: Product Hunt reviewers report that the UI becomes cumbersome with large datasets and long experiment histories, and filtered-result sharing is awkward. · producthunt.com, 15 Oct 2026
Limit: learning curve: Reviewers commonly describe the interface and concepts as dense or overwhelming for new users. · g2.com, 15 Oct 2026
Limit: outage: On May 1, 2025, a certificate and DNS configuration error caused a 28-minute US API, web and SDK outage with about 55% of requests failing. · blog.langchain.com, 15 Oct 2026