# LangSmith

> Source: https://agentsindex.ai/langsmith

LangChain's framework-agnostic platform for tracing, evaluating, monitoring and deploying LLM agents.

- Best for: AI teams who need to trace, evaluate, and ship agents with production visibility.

Starting price

Free Developer plan

Paid plans from **$39 a seat a month**

From [langchain.com/pricing](https://www.langchain.com/pricing), checked 18 November 2026

Paid plans in Agent tools and integrations

**#26** cheapest of 46, tied with 2

**$30** median a month

Cheapest $3, median $30 Priciest $650

- **25** cheaper
- **2** same price
- **18** pricier

Pricier than 25 of the other 45 with a monthly price in agent tools and integrations.

Vendor pricing pages as of 20 November 2026 n = 46 with a monthly price [All 98 in agent tools and integrations](https://agentsindex.ai/categories/agent-tools-and-integrations)

- MCP server Yes From langchain.com/pricing, 18 September 2026 [langchain.com](https://www.langchain.com/pricing)
- API Yes From langchain.com/pricing, 18 September 2026 [langchain.com](https://www.langchain.com/pricing)
- Open source Partly From langchain.com/langchain, 18 September 2026 [langchain.com](https://www.langchain.com/langchain)
- Self-host Yes

[Visit LangSmith](https://agentsindex.ai/go/langsmith) [Compare](https://agentsindex.ai/compare?cat=agent-tools-and-integrations&products=langsmith)

What we checked

- Pages last read: 18 November 2026
- Facts checked: 5 facts · From 2 sources, read 10 November 2026
- AI answers: Too few read · 58 buyer-question answers read, November 2026

[How we check](https://agentsindex.ai/methodology) [Editorial policy](https://agentsindex.ai/editorial-policy#not-for-sale)

## At a glance

- Category: [Agent tools and integrations](https://agentsindex.ai/categories/agent-tools-and-integrations)
- Tags: Freemium, Open source, Self-hosted

### For developers

- Works with: OpenTelemetry, A2A, MCP, Agent Protocol, Pagerduty, OpenAI and 5 more

## What is LangSmith?

LangSmith is an AI observability and evaluation platform for teams that need to trace, debug, score, and deploy agents. It combines Observability, Evaluation, Deployment, Prompt Hub, Playground, and Canvas, plus annotation queues for human feedback. LangSmith works with Python, Typescript, Go, or Java SDKs and is used by Klarna, Vanta, monday.com, and Cloudflare. Plans run Developer $0 / seat per month, Plus $39 / seat per month, and Enterprise custom pricing.

## LangSmith is now six metered products under one name

LangSmith started as LangChain's tracing tool, and the [pricing page](https://www.langchain.com/pricing) now bills six separate services under the name: Observability & Evaluation, Agent Server, Fleet, Engine, Sandboxes and an LLM Gateway. Each has its own meter and its own free allowance, so a comparison article that prices "LangSmith" usually prices only the tracing part.

The core is still [observability](https://www.langchain.com/langsmith/observability): step-by-step agent traces, dashboards for "token usage, latency (P50, P99), error rates, cost breakdowns, and feedback scores", online LLM-as-judge and code evals, and alerts through webhooks or PagerDuty. Around it sit Agent Server for hosting long-running agents (with "30+ API endpoints including state and memory" and the option to "Expose agent as MCP server"), Fleet for building agents "using everyday language, not code", Sandboxes for running agent-generated code, and the LLM Gateway for cost controls, model fallbacks and "PII and secrets redaction".

Engine is the newest and the most agent-like piece. The pricing FAQ says it monitors traces, clusters failures into issues and recommends fixes to prompts and code, and that it "is scheduled to run once every 6 hours". An organization admin can switch it off in Settings, after which it stops consuming usage.

## It traces agents that never touch LangChain

The most common belief about LangSmith is that it requires the LangChain framework. The vendor's own [observability FAQ](https://www.langchain.com/langsmith/observability) says otherwise: "LangSmith works with any LLM framework. Trace applications built with OpenAI SDK, Anthropic SDK, Vercel AI SDK, LlamaIndex, or custom implementations, not just LangChain." It ships SDKs for Python, TypeScript, Go and Java, and supports OpenTelemetry in both directions: "Send LangSmith trace data to your tools or ingest OTel data into LangSmith."

The same page states that tracing runs off the request path: "The LangSmith SDK uses an async callback handler that sends traces to a distributed collector", and "If LangSmith experiences an incident, your agent keeps running normally." Observability and evaluation can be adopted separately, with tracing first and evals later.

Getting traces flowing still means instrumenting code with an SDK or an OTel exporter, so a developer is needed for setup.

## Seats are the small part of the bill; usage is the rest

The [pricing page](https://www.langchain.com/pricing) shows three plans. Developer is free for a single seat with 5k base traces a month. Plus is "$39 / seat per month then pay as you go" with 10k base traces a month and unlimited seats. Enterprise is "Custom pricing", invoiced annually upfront. Developer and Plus bill "Monthly, self-serve", and no annual toggle is offered for them.

Everything past the included allowances is charged in LangChain Standard Units, and the page states "1 LSU = $1". Additional traces cost 0.005 LSU each. Agent Server runtime is metered by vCPU-hour and GiB-hour, Sandboxes are "billed per second", Tuned Evaluators run at 0.015 LSU per evaluation run, and the FAQ estimates that one Engine run "can consume ~7-45 LSUs", which at four scheduled runs a day adds up fast. The FAQ also says that on Plus "LSUs are an additional charge on top of your monthly license fee".

The free allowances differ by product. Developer includes 7 LSU of Fleet a month, after which "Fleet pauses new runs until the allowance resets or the organization upgrades to Plus"; Plus includes 37 LSU and bills the excess. Both plans include 8 LSU of Sandbox usage, and Developer is "Limited to 10 Sandboxes". Plus adds "1 Small Agent Server included on serverless hosting at no charge". Developer has no Agent Server or Engine access at all.

A usage calculator sits on the same page and marks itself "Estimate only. Actual LSU consumption varies with the work your agents perform."

## Extended trace retention dropped to 180 days in September 2026

Traces come in two tiers. Base traces are kept 14 days; extended traces cost more and are kept longer. The [billing docs](https://docs.langchain.com/langsmith/billing) record a change: "Starting September 14, 2026, the maximum long-lived trace retention period for SaaS customers is changing to 180 days." Enterprise customers can tune the extended period per workspace only up to that ceiling, and the change applies to new traces only. The [pricing FAQ](https://www.langchain.com/pricing) now gives the same 180 days, so older reviews quoting a longer window are out of date.

Retention is set per tracing project, and automation rules can promote a sample of traces (for example "10% of all traces" or "Errored traces") to the extended tier. When a rule matches any run in a trace, the whole trace is upgraded. The docs warn that "The extended data retention limit can cause features other than tracing to stop working once reached", so evaluators and rules that extend retention are the parts to budget for.

Spend is capped per workspace: a spend limit is converted into base and extended trace limits, with a slider to split them. In multi-workspace organizations the workspace limits are approximate and "the organization-level spend limit is absolute".

## Self-hosting is an Enterprise add-on, described two ways

The [self-hosting docs](https://docs.langchain.com/langsmith/self-hosted) open with the condition: "Self-hosted LangSmith is an add-on to the Enterprise plan designed for LangChain's largest, most security-conscious customers." The base install covers observability, evaluation, prompt engineering, Insights and Chat; Sandboxes, Engine, Fleet, the LLM Gateway (beta) and Agent Server hosting are optional modules, each with its own requirements. The [pricing table](https://www.langchain.com/pricing) agrees: Developer and Plus are Cloud only, Enterprise is "Cloud, Hybrid, or Self-Hosted".

The [observability page](https://www.langchain.com/langsmith/observability) is looser. It answers "Can I self-host LangSmith?" with a plain "Yes" and lists "managed cloud, bring-your-own-cloud (BYOC), and self-hosted options", naming BYOC where the pricing table names Hybrid. It also promotes self-hosting SmithDB, the new trace store, "inside your VPC" with "three stateless components on object storage and Postgres".

Data location is stated two ways. The observability FAQ says data on smith.langchain.com "is stored in GCP us-central-1", while the pricing table's Data location row reads "LangChain's Cloud (US or EU)" for Plus and Enterprise. Both pages are current, and the EU option appears only on the second.

## The trust center publishes a full audit pack

On training, both the [pricing FAQ](https://www.langchain.com/pricing) and the observability page say the same thing: "LangSmith does not use your data to train models." Custom SSO, ABAC and RBAC are Enterprise items; Developer and Plus sign in with Google or GitHub. A support SLA appears only on the Enterprise card.

## Where it fits, and where it does not

LangSmith suits engineering teams that want tracing, evals and agent hosting from one vendor, especially teams already on LangGraph, where Agent Server and Studio connect directly. Its framework-agnostic SDKs and OTel support make it a fair choice for teams on other stacks too, as the [observability FAQ](https://www.langchain.com/langsmith/observability) states.

It fits less well where cost has to be fixed in advance. The seat price is small next to usage across traces, Agent Server compute, Sandboxes, Fleet, Engine and Gateway Credits, and the vendor's own calculator calls its figures estimates. Teams that need traces kept beyond 180 days on the cloud service now have to export them or move them into datasets. Buyers who must keep data in their own network need the Enterprise contract, since the [self-hosting docs](https://docs.langchain.com/langsmith/self-hosted) make it an add-on to that plan.

Our take

Developer includes up to 5k base traces per month, while Plus raises that to 10k and adds unlimited seats. If your trace volume is near either ceiling, verify how quickly pay-as-you-go charges would begin.

## How much does LangSmith cost?

Prices excluding tax, as LangSmith publishes them.

- Developer · $0/seat per month
  - Up to 5k base traces / mo, then pay-as-you-go
  - Tracing to debug agent execution
  - Online and offline evals
  - Prompt Hub, Playground and Canvas for auto improving prompts
  - Annotation queues for human feedback
  - Monitoring and alerting
  - 1 Fleet agent
  - Community support
  - 1 seat
- Plus · $39/seat per month
  - Everything in the Developer plan, and:
  - Up to 10k base traces / mo, then pay-as-you-go
  - 1 dev-sized agent deployment included
  - Email support
  - Unlimited Fleet agents
  - Add unlimited seats
  - Up to 3 workspaces
- Enterprise · Custom pricing
  - Everything in the Plus plan, and:
  - Alternative hosting options, including hybrid and self-hosted so data doesn't leave your VPC
  - Custom SSO and RBAC
  - Acccess to deployed engineering team
  - Support SLA
  - Team trainings & architectural guidance
  - Custom Fleet packages

[![LangSmith](https://agentsindex.ai/img/1024/screenshots/langsmith/pricing.webp)](https://agentsindex.ai/img/1920/screenshots/langsmith/pricing.webp)

Pricing page, captured 19 October 2026

[langchain.com/pricing](https://www.langchain.com/pricing), checked 18 November 2026

### All plans and limits

| Plan | Price | Base traces | Fleet LSUs | Sandboxes |
| --- | --- | --- | --- | --- |
| Developer | $0 per seat a month | Up to 5k / mo, then pay-as-you-go | 7 LSUs included per organization each month | Includes 8 LSU Limited to 10. |
| Plus | $39 per seat a month | Up to 10k / mo, then pay-as-you-go | 37 LSUs included per organization each month | Includes 8 LSU of Sandbox usage / mo for free. |
| Enterprise | Price on request | Not stated | Not stated | Not stated |

Read 18 November 2026 from [langchain.com/pricing](https://www.langchain.com/pricing).

## For AI agents

- MCP server: Yes [langchain.com, 18 September 2026](https://www.langchain.com/pricing)
- API: Yes [langchain.com, 18 September 2026](https://www.langchain.com/pricing)

## What we checked on LangSmith

Yes Partly No

What it costs to start Yes

### What the price counts

Priced per seat plus usage: Plus costs $39 per seat a month with 10k base traces included, then pay-as-you-go trace and LSU metering; Enterprise is quoted.

> $39 / seat per month then pay as you go

Plan cards, Plus column price [langchain.com/pricing](https://www.langchain.com/pricing) Checked 10 November 2026

What it costs to start Yes

### Free to use

A standing $0 Developer plan is self-serve with no sales call: one seat, 5k base traces a month, community support, then pay-as-you-go beyond the included usage.

> Developer — For solo users getting started. $0 / seat per month... Up to 5k base traces / mo, then pay-as-you-go. Community support. 1 seat

Plan cards, Developer column [langchain.com/pricing](https://www.langchain.com/pricing) Checked 24 October 2026

How you build with it Yes

### Public API

The LangSmith REST API is open to every plan and Fleet agents can be triggered by API, but the Agent Server's assistants and state endpoints arrive with the Plus plan.

> 30+ API endpoints including state and memory

Plan comparison table, LangSmith Deployment section [langchain.com/pricing](https://www.langchain.com/pricing) Checked 18 September 2026

How you build with it Yes

### MCP server

Agents can be exposed as MCP servers on every plan including the free one, and can themselves consume tools from remote MCP servers; LangSmith also publishes its own MCP server.

> Expose agent as MCP server

Plan comparison table, LangSmith Deployment section [langchain.com/pricing](https://www.langchain.com/pricing) Checked 18 September 2026

How you build with it Partly

### Open source

The frameworks and the dcode coding agent are MIT-licensed and free, while the LangSmith platform that carries the price, seats and governance is proprietary and hosted.

> Yes - LangChain is an MIT-licensed open-source library and is free to use.

FAQs for LangChain, 'Is LangChain open source?' [langchain.com/langchain](https://www.langchain.com/langchain) Checked 18 September 2026

## Where we've written about LangSmith

- [AI agent pricing benchmark: what MCP-ready tools charge](https://agentsindex.ai/blog/ai-agent-pricing-benchmark)

## Questions about LangSmith

- How much does LangSmith cost?: LangSmith's Developer plan is free. Its paid plans are Plus at $39 per seat a month and Enterprise. From langchain.com/pricing, checked 18 November 2026.
- Does LangSmith have an API or an MCP server?: LangSmith has an API. From langchain.com/pricing, checked 18 September 2026. LangSmith has an MCP server. From langchain.com/pricing, checked 18 September 2026.
- Is LangSmith open source, and can you host it yourself?: Part of LangSmith is open source. From langchain.com/langchain, checked 18 September 2026. You can host LangSmith yourself.
- What have we checked on LangSmith?: 5 facts, under these headings: What it costs to start, how you build with it.
- How do we check LangSmith?: We hold 5 facts on LangSmith, quoted from 2 pages. Facts last checked 10 November 2026. Price last checked 18 November 2026. Last change found 21 March 2026.

[How we check every listing](https://agentsindex.ai/methodology)

## Sources and changes

Last changed 21 March 2026

Sources on langchain.com 2

- Read 18 November 2026 · [/pricing](https://www.langchain.com/pricing)
- Read 18 September 2026 · [/langchain](https://www.langchain.com/langchain)

## LangSmith price history

Plus has cost $39 a month since 5 August 2025, on all 11 readings.

| Price | Plan | First seen | Last seen | Readings | Source |
| --- | --- | --- | --- | --- | --- |
| $39 a month | Plus | 5 August 2025 | 20 November 2026 | 11 | [Archived copy, 5 August 2025](https://web.archive.org/web/20250805044729/https://www.langchain.com/pricing) |

From 11 readings of LangSmith's pricing page: 6 Internet Archive copies from 5 August 2025 and 5 of our own checks since 31 October 2026. A date is when we read the page, not when the price changed.

## Similar in Agent tools and integrations

[See all in Agent tools and integrations](https://agentsindex.ai/categories/agent-tools-and-integrations)

- [ERIUS PHONE](https://agentsindex.ai/eriusphone) Rented cloud Android phones, each in its own container, that keep an agent's apps and sign-ins between sessions. From about $44 a month
- [TinyFish](https://agentsindex.ai/tinyfish) Web search, page fetching, cloud browsers and goal-driven web automation for AI agents, with search and fetch free. Free plan, then $0.002 per minute
- [Smithery](https://agentsindex.ai/smithery) Public MCP server registry, now run by Arcade, listing direct vendor MCP URLs after its proxy shut down. Free plan
- [UI Discovery](https://agentsindex.ai/uidiscovery) A library of real websites, iOS apps and dashboards with each one's palette, fonts and original build prompt. From $20 a month

## From our research

[All research](https://agentsindex.ai/blog)

- [Comparison](https://agentsindex.ai/blog/ai-coding-agents-what-each-is-good-at) 14 Aug 2026 AI coding agents compared: what each one is actually good at Every major coding agent now costs $20. We compare what that money actually buys, which projects are still maintained, and which tool fits which job.
- [Comparison](https://agentsindex.ai/blog/ai-customer-service-agents-cost-per-resolution) 13 Aug 2026 AI support agents: what each costs per resolved chat Published prices for 21 AI support agents, converted to one comparable number: cost per conversation the AI closed on its own. Verified 13 Aug 2026.
- [Buyer's guide](https://agentsindex.ai/blog/ai-design-agents-for-teams-without-a-designer) 4 Oct 2026 AI design agents for teams without a designer Gamma vs. Beautiful.ai for decks, and three ad-creative generators compared by cost per asset.

## Is LangSmith yours?

Free, and the page then carries the verified owner mark. You confirm it with an email at langchain.com. [Claim this listing](https://agentsindex.ai/claim/langsmith?from=owner-line)

[Report an error](mailto:hello@agentsindex.ai?subject=Report%20an%20error%3A%20LangSmith%20(langsmith)&body=What%20is%20wrong%20on%20the%20page%2C%20and%20where%20on%20the%20product) if a fact is wrong.
