Research

LangSmith Alternatives: Who Should Stay, and Where a Prompt Test Fits

Last checked 18 November 2026.

LangSmith is LangChain's platform for tracing, testing and monitoring LLM apps. Teams look at alternatives for different reasons: per-seat pricing as the team grows, a 14-day limit on base traces, or wanting a quicker way to compare models than writing an experiment. This guide sets out what LangSmith does better, who should stay with it, where Spring Prompt fits, and how to bring a dataset across.

Have a LangSmith dataset? Load the export into a Prompt test and read every model's answers side by side. Test your prompt →

LangSmith's plans, as of November 2026

PlanPriceIncludedHosting
Developer$01 seat; up to 5,000 base traces a month, then pay as you goCloud
Plus$39 per seat a monthUnlimited seats; up to 10,000 base traces a month, then pay as you goCloud
EnterpriseCustomCustom seats and workspacesCloud, hybrid or self-hosted

Base traces are kept for 14 days and extended traces for 180 days. Usage beyond the allowance is metered in LSUs, at $1.00 each. Self-hosted LangSmith is an add-on to the Enterprise plan, licensed with a key from LangChain.

Sources: LangChain pricing and LangSmith's self-hosting docs, both checked 18 November 2026.

What LangSmith does better

  • Tracing. LangSmith records traces from your running app, so you can see each step of a chain or agent. Spring Prompt doesn't trace. Its Live package samples real calls to build tests, not to monitor production.
  • Versioned datasets. Its dataset docs say every change to a dataset's examples creates a new version, and you can tag one, such as “prod”, and evaluate against it.
  • Tests in your test runner. The pytest plugin turns marked tests into dataset examples and each run into an experiment, and the JavaScript SDK has a similar Vitest and Jest integration. That's a natural fit if your evals already live in your test suite.
  • Your own infrastructure. On Enterprise you can self-host, or keep the data plane in your own cloud with LangChain running the control plane. Spring Prompt is hosted only.
  • Scale. Experiments run over whole datasets. A Prompt test is capped at 24 cases.

Who should stay with LangSmith

If you build with LangChain or LangGraph, rely on traces to debug agents, or keep evals in pytest or Vitest, LangSmith does more of that than we do. Stay with it. A Prompt test is useful alongside it when you want to compare models on one prompt quickly, or show the results to someone outside the engineering team.

Where Spring Prompt fits

  • Prompt test. Paste a prompt in a browser. We draft test cases and checks for you to edit, run the models you pick side by side, check every answer against your rules, and show the cost and time of each answer next to whether it passed. No SDK, no per-seat price.
  • Live. A two-line package, springprompt on PyPI and npm, that copies a sample of your app's real model calls so you can test them on other models. See the Live docs.
  • Check before merge. incremento/springprompt-action@v1 runs a changed prompt or model on a saved test and returns pass, review or fail. See how it works.
  • Independent benchmarks. Our benchmarks score current models on business tasks and show their real answers, so you can choose which models to test.

LangSmith vs Spring Prompt

LangSmithSpring Prompt
Main jobTracing, datasets and experiments for LLM appsTesting a prompt across models
SetupSDK in your app or test suitePaste a prompt; optional two-line Live package
Writing testsDatasets and evaluators you buildWe draft cases and checks from your prompt; you edit them
Test sizeWhole datasetsUp to 24 cases and 10 checks
Production tracesYes; base traces kept 14 daysNo; Live samples calls for testing
CIpytest, Vitest and Jest integrationsincremento/springprompt-action@v1 on pull requests
HostingCloud; hybrid or self-hosted on EnterpriseHosted only
PricingFree for one seat, Plus $39 per seat a month, Enterprise customPay for the tests you run; free trial credit

Get your data out of LangSmith

  • Datasets. In the Dataset & Experiments tab, select a dataset, click ⋮ at the top right, then Download Dataset. It comes as CSV, JSONL or OpenAI's fine-tuning format. If you use version tags, note which version is the one you trust.
  • Prompts. Copy out the full text of each prompt you run in production, including the system message.
  • Evaluators. Write down what each one checked and how. A judge's instructions are worth keeping word for word.

What loads into a Prompt test

In a new Prompt test, choose Import from my app, then paste or upload the file. These load with no model call:

You haveWhat comes in
A LangSmith JSONL or CSV export (inputs and outputs)Each example as a case. Named inputs become {{variables}}; inputs holding messages come in as a chat, with the system message kept. Outputs become each case's expected answer, with a judged check that the answer matches it.
The OpenAI fine-tuning export ({"messages": [...]})The conversation as the input, and the last assistant reply as the expected answer.

A dataset holds inputs, not your prompt, so when there isn't one the import makes a stand-in from the fields and tells you. Paste your real prompt over it, keeping the {{variables}}. Each example's metadata is left out, and the import says so.

A Prompt test takes up to 24 cases and 10 checks, so a large dataset is cut to the first ones. Pick the examples that matter most: the awkward inputs, not the easy ones. API keys in the file are removed before anything is stored.

Other guides: Langfuse, Braintrust, Promptfoo and Humanloop.

Bring your dataset, or start from a prompt. Run the models you're choosing between on the same inputs and read every answer side by side. Test your prompt →

Quick answers

How much does LangSmith cost?

LangChain's pricing page lists Developer at $0 for one seat with up to 5,000 base traces a month, Plus at $39 per seat a month with up to 10,000 base traces a month, and Enterprise at custom pricing. Usage beyond that is pay as you go. We checked on 18 November 2026.

Can I self-host LangSmith?

Only on Enterprise. LangSmith's docs say self-hosted LangSmith is an add-on to the Enterprise plan, and the pricing page lists cloud, hybrid and self-hosted deployment for Enterprise only.

How long does LangSmith keep traces?

According to the pricing page, base traces are kept for 14 days and extended traces for 180 days. Base traces can be upgraded to extended for a fee.

How do I export a LangSmith dataset?

In the Dataset & Experiments tab, select the dataset, click the three-dot menu at the top right and choose Download Dataset. It comes as CSV, JSONL or OpenAI's fine-tuning format.

Can I import a LangSmith dataset into Spring Prompt?

Yes. The JSONL or CSV export loads into a Prompt test without a model call. Inputs become variables, or a chat when they hold messages, and outputs become each case's expected answer. A Prompt test takes up to 24 cases and 10 checks.

More research
All posts →