LangSmith pricing - what you pay for, how traces are billed, and how to estimate monthly cost

LangSmith pricing has two main parts: seats (who can use the workspace) and trace usage (how many application runs you store). Retention tiers matter because they change the price per trace. This page is built to be fully responsive for mobile and desktop, with tables, FAQs, and a spend calculator.

Developer: $0/seat (max 1 seat), includes 5k base traces/mo Plus: $39/seat/mo, includes 10k base traces/mo Base retention: 14 days • Extended retention: 400 days Base: $2.50/1k • Extended: $5.00/1k • Upgrade: $2.50/1k

1) Overview: what “LangSmith pricing” includes

People search “LangSmith pricing” for one of three reasons: (1) they want the cheapest plan to start; (2) they need to understand how traces are billed; or (3) they want to avoid surprise bills when production traffic scales. The reality is that LangSmith pricing is not a single meter.

Your monthly cost can include:

Seat fees (Plus): a predictable cost based on how many users access the workspace.
Trace usage charges: after you exceed included base traces, you pay per 1,000 traces, with different rates by retention tier.
Retention decisions: base traces (short-lived) vs extended traces (long-lived) can change cost meaningfully at scale.
Plan-dependent extras: deployment and support options differ by plan and can influence total value.

The best way to budget is: (seats) + (billable base traces × base rate) + (extended traces × extended rate), then set limits.

What is a “trace”?

A trace represents one end-to-end execution of your application (agent run, chain run, evaluator run, or playground session). A single trace can contain many internal steps such as LLM calls and tool calls. Tracing everything helps debugging and evaluation, but it also means your trace count can grow with product usage.

2) Plans: Developer vs Plus vs Enterprise

LangSmith typically offers a free Developer plan for solo builders, Plus for teams, and Enterprise for organizations that need advanced security, compliance, hosting, and SLAs.

Plan	Price	Seats	Included base traces / month	Best for
Developer	$0/seat/month	Max 1 seat	5,000 base traces	Solo prototyping, debugging, early evals
Plus	$39/seat/month	Unlimited seats	10,000 base traces	Teams, collaboration, higher volume tracing
Enterprise	Custom	Custom	Custom	SSO/RBAC, advanced support, custom hosting

How to choose fast

Choose Developer if you are one person and your usage is mostly local/dev with moderate trace volume.
Choose Plus if multiple people need access and you want predictable collaboration + more included traces.
Choose Enterprise if compliance/security/hosting is the main requirement, not just price.

3) Trace pricing: base vs extended retention

Retention is the most important pricing knob after seat count. LangSmith distinguishes: base traces (short retention) and extended traces (long retention). Teams often start by keeping everything, then later adopt policies like “keep only high-signal traces for long retention.”

Tier	Retention	Common purpose	Price (per 1,000 traces)
Base	14 days	Debugging, quick iteration, short-term monitoring	$2.50
Extended	400 days	Long-term improvement, feedback, labeling, governance	$5.00
Upgrade base → extended	—	Keep selected traces longer after capture	$2.50

Why retention changes cost

Short retention is designed for high-volume debugging data that loses value quickly. Long retention is more expensive because the traces remain searchable and usable for long-term evaluation, product analytics, and governance workflows.

How trace volume grows in real products

Trace counts often scale with:

Requests per day × how many environments you trace (dev + staging + prod).
Sampling strategy (100% capture vs partial sampling for success cases).
Agent complexity (more steps doesn’t necessarily mean more traces, but it can increase payload and “events”).

A common approach is: keep 100% of error traces and a sampled subset of successful traces, and upgrade to extended retention only for traces that include feedback or become “golden examples.”

4) Billing: when you’re charged and how to avoid surprises

Billing usually breaks down into:

Seats: predictable monthly subscription (Plus).
Traces: usage-based monthly charges, influenced by included allowances and retention tier.

Practical spend controls

Set usage limits: cap monthly trace ingestion so your bill cannot exceed a threshold.
Track draft invoices: check usage weekly during launches or heavy testing.
Use retention intentionally: base by default, extended only for high-value traces.
Sample in production: do not store every successful request if you don’t need it.

Recommended “starter policy”

Starter policy: base retention for everything; extended retention only for (a) user-feedback traces, (b) incident-related traces, and (c) curated dataset examples. Sampling for non-error traffic at 10–20%.

5) Monthly cost calculator

This calculator is built into the page and works offline. Adjust inputs and it will estimate monthly total based on seats + traces. If your account has different rates or promotions, override the defaults in “Advanced.”

Inputs

Plan

Seats

Base traces (14-day retention)

Extended traces (400-day retention)

Advanced (override defaults)

Plus seat price ($/seat/month)

Developer included base traces

Plus included base traces

Base rate ($ per 1,000 traces)

Extended rate ($ per 1,000 traces)

Upgrade base→extended ($ per 1,000)

Tip: If you want to model “some traces are upgraded,” subtract those from base traces and add them to extended traces.

Estimated monthly total

$0.00

6) Cost examples (quick scenarios)

Example 1: Developer plan, 4,000 base traces

You’re under the 5,000 included base traces. Total stays at approximately $0 (excluding any add-ons), making Developer a great option for personal projects and early experiments.

Example 2: Plus plan, 5 seats, 60,000 base traces, 5,000 extended traces

Seat cost = 5 × $39 = $195/month. Included base traces = 10,000. Billable base traces = 60,000 − 10,000 = 50,000 → 50 × $2.50 = $125. Extended traces = 5,000 → 5 × $5.00 = $25. Estimated total ≈ $195 + $125 + $25 = $345/month.

Example 3: Heavy production tracing with long retention everywhere

If you store most traces as extended at high volume, cost scales faster. In high-traffic apps, the best control is a retention policy + sampling strategy so only high-signal traces live long-term.

Rule of thumb: If you’re not sure what to keep long-term, start with base retention + sampling. Then “upgrade to extended” for the traces you repeatedly reference for debugging, evaluation, or feedback.

7) LangSmith self-hosted pricing

LangSmith self-hosted pricing typically refers to the Enterprise option where LangSmith can be deployed in your own environment (or a controlled/private setup) for organizations that need strict data residency, compliance, and security controls. Pricing is usually custom because it depends on your deployment model (cloud, VPC, on-prem), required security features (SSO/RBAC), audit/compliance needs, support SLAs, and expected usage scale. In practice, “self-hosted” is less about a public price table and more about a packaged agreement that includes implementation guidance and enterprise support.

8) LangSmith enterprise pricing

LangSmith Enterprise pricing is designed for teams that require more than standard self-serve plans especially around governance and reliability. Enterprise agreements are commonly tailored to your organization’s needs: user access at scale, role based permissions, SSO, advanced support, contract terms, data controls, and sometimes dedicated infrastructure or custom hosting. If you’re evaluating Enterprise, the most important pricing variables are your security/compliance requirements, expected trace volume/retention strategy, and the level of support/SLAs your production workloads require.

9) LangSmith API pricing

LangSmith API pricing is usually not “per API call” in the way model providers charge for tokens. Instead, LangSmith’s “API usage” cost is commonly tied to observability usage especially how many traces your application sends (and how long you retain them). The practical way to think about LangSmith API cost is: how much telemetry you ingest from your apps (agents, chains, evaluations) and whether those traces are stored with short or extended retention. To keep costs predictable, teams often use sampling, trace filtering, and usage limits capturing 100% of errors and only a portion of successful requests.

10) LangSmith features pricing

LangSmith features pricing varies by plan, with advanced capabilities usually unlocked at higher tiers. The free/Developer tier is often best for individual builders needing core tracing and basic evaluation workflows. The Plus tier typically adds paid seats, higher included usage, and team centric features for collaboration and production readiness. Enterprise tiers generally unlock the full set of organizational features SSO/RBAC, stronger admin controls, compliance options, SLAs, and potentially self-hosted/hybrid deployments. When comparing feature pricing, focus on what matters for your workflow: team seats, trace volume, retention needs, evaluation scale, and governance/security requirements.

Quick checklist

If you do these five things, LangSmith costs become predictable.

Pick the plan by seats (Developer solo; Plus team).
Default to base retention for most traces.
Use extended retention only for high-signal traces.
Sample production traffic; capture 100% errors.
Set monthly usage limits and check usage weekly.

Best starting point: base retention + 10% sampling in prod + keep errors. Upgrade to extended only for feedback traces and “golden examples.”

Mini glossary

Seat: one user with workspace access.
Trace: one end-to-end application run.
Base: 14-day retention (cheaper).
Extended: 400-day retention (more expensive).

Mobile UX note

This page includes: a sticky header, a mobile menu, responsive grid layout, scrollable tables, and accessible controls (labels + aria).

7) FAQ

What is LangSmith pricing based on?

LangSmith pricing typically combines seat-based pricing (who can access the workspace) and usage-based pricing (how many traces you store), with costs influenced by retention tier (base vs extended).

Is LangSmith free?

Yes. The Developer plan is designed as the free tier for one seat, with an included monthly trace allowance.

What is the LangSmith Developer plan?

Developer is the entry plan for solo builders. It typically includes 1 seat and an allowance of base traces for tracing, debugging, and basic eval workflows.

What is the LangSmith Plus plan?

Plus is the paid plan for teams. It usually supports multiple seats, provides higher included usage, and adds team-focused features like collaboration and support.

What is LangSmith Enterprise pricing?

Enterprise pricing is custom and typically depends on security (SSO/RBAC), hosting needs, support SLAs, compliance requirements, and usage scale.

What is a “seat” in LangSmith pricing?

A seat is a user who has access to the LangSmith workspace UI and features (projects, traces, datasets, evaluations, admin controls).

Do I pay per user or per workspace?

Self-serve plans usually bill per seat plus usage (traces beyond allowance). Enterprise may be a negotiated package.

What is a “trace” in LangSmith billing?

A trace is generally one end-to-end run of your app (agent, chain, evaluator, or playground execution), containing multiple steps/spans.

Do I pay for each LLM call?

Pricing is typically structured around traces and retention, not directly per LLM call. However, more complex runs can create larger trace payloads and more internal events.

What is “base retention”?

Base retention means traces are stored for a short period (commonly used for debugging). After the retention window, traces expire/delete.

What is “extended retention”?

Extended retention stores traces for a long period, useful for long-term evaluation, feedback tracking, and regression analysis over time.

Why does retention affect pricing?

Longer retention increases storage/search/governance overhead and long-horizon utility—so extended retention is typically priced higher than base.

What’s the difference between base traces and extended traces?

They are the same “trace” concept but stored in different retention tiers: base (short-lived) vs extended (long-lived).

Can I mix base and extended retention in one workspace?

Yes. Many teams default to base retention and upgrade only high-signal traces (feedback, incidents, golden examples) to extended retention.

How do I choose what to keep as extended?

Keep extended traces for user feedback, critical failures, release/regression checks, incident investigations, and curated benchmark examples.

What causes costs to spike unexpectedly?

Common causes include tracing 100% of production traffic, defaulting to extended retention, load tests, launches, or not setting usage limits.

Do evaluations add to trace usage?

Yes. Evaluation runs usually generate traces. Frequent evals on large datasets can increase monthly trace counts significantly.

What are “ingested events”?

Ingested events are the smaller telemetry pieces within traces (spans/start/end/feedback events). Budgeting is usually easiest by trace count.

Is trace volume the same as user traffic?

Not always. It depends on your sampling, environments traced, internal jobs, and whether you trace every request end-to-end.

Does tracing dev/staging count?

If you send traces to LangSmith, they generally count regardless of environment. Many teams trace 100% in dev/staging but sample in production.

How often am I billed?

Billing is typically monthly. Seats follow subscription billing; usage (traces) is calculated for the month’s total usage.

Are seats prorated if I add users mid-month?

Many systems prorate seat costs when you add seats partway through a billing period. Confirm your account’s billing behavior in the billing page.

Can I set a monthly cap?

Yes. Use usage limits/caps to control how many traces can be ingested per month and to prevent accidental overspend.

What happens when I hit a trace limit?

Depending on configuration, new traces may stop ingesting or you may need to increase limits for additional tracing during the same month.

Can I keep tracing errors while sampling success traffic?

Yes. A best practice is 100% capture for error traces plus a sampling rate for successful requests.

What’s a good production sampling rate?

Many teams start around 5–20% for successful requests and adjust up/down based on reliability needs and budget.

Should I trace everything in production?

Usually no. Most teams use selective tracing: all errors, sampled successes, and special cohorts (new releases, specific customers, beta features).

How do I estimate trace volume?

Estimate requests/day × days/month × tracing coverage %, then add evaluation/batch-job traces. Split into base vs extended retention.

How do I reduce costs without losing observability?

Default to base retention, sample success traffic, keep 100% errors, upgrade only high-signal traces to extended, and set usage limits.

Is LangSmith cost separate from model token costs?

Yes. LangSmith covers observability/evaluation. Your LLM provider bills tokens separately.

Do tool calls increase my bill?

Tool calls increase trace complexity and telemetry events, but the main cost driver is still how many end-to-end traces you ingest and retain.

Does the number of spans in a trace matter?

Not usually as a direct unit price, but very verbose traces can increase payload and operational overhead—so keep instrumentation purposeful.

What does “upgrade base → extended” mean?

It means you can capture traces with short retention first, then convert selected high-value traces to extended retention for long-term keeping.

When should I upgrade a trace to extended?

Upgrade traces linked to user feedback, regressions, postmortems, release validation, and curated benchmark examples.

Best retention policy for early-stage projects?

Start with base retention for everything. Add extended only when you discover stable “golden examples” and feedback loops you want to keep long-term.

Best retention policy for mature production systems?

Base retention for general traffic, extended for feedback and golden traces, and temporary extended retention during incident or release windows.

Can I delete traces to reduce costs?

Manual deletion may help governance, but billing usually depends on ingested usage and retention tier, not only current stored volume.

Is there startup pricing or discounts?

Many platforms offer startup programs at times. Check the official pricing page/programs for current availability and eligibility.

Is there annual billing?

Enterprise often supports annual terms; self-serve plans are commonly monthly. Confirm in your billing settings.

Can I use multiple workspaces to separate costs?

Yes. Teams often separate dev vs prod workspaces, but be mindful of operational overhead and duplicated settings.

How do I find what generates the most traces?

Tag traces with metadata like service name, environment, version, route, or customer segment so you can filter and attribute volume.

Does LangSmith provide usage dashboards?

Most setups include usage/invoice views. Use them as a weekly habit during scaling and launches.

What’s the cheapest team setup?

Use the minimum seats you need, base retention by default, production sampling for success traffic, and extended retention only for high-signal traces.

Do contractors require paid seats?

If they need workspace access, they typically use a seat. Common practice: time-box access and remove seats when work ends.

What justifies Plus most often?

Collaboration, higher included usage, shared evaluation workflows, and team support—especially when multiple engineers are shipping to production.

What justifies Enterprise most often?

SSO/RBAC, compliance demands, special hosting needs, SLAs, and dedicated support for large-scale production workloads.

How do I budget evaluation at scale?

Estimate dataset size × eval frequency × variants (models/prompts). Keep most eval traces base unless you need long-term raw trace detail.

Can shorter retention reduce compliance risk?

Yes. Short retention reduces the amount of stored data and can lower governance overhead, while often reducing cost as well.

What’s the simplest “safe setup” for predictable bills?

Base retention by default, 10% sampling in production, 100% error traces, upgrade only feedback/golden traces to extended, and set usage limits.

What’s the #1 mistake teams make with LangSmith pricing?

Turning on 100% production tracing with extended retention by default—then realizing the trace volume is far higher than expected.

Do I pay for every tool call or LLM call?

LangSmith usage is typically reasoned about in terms of end-to-end traces per month. However, a single trace can contain many internal steps and events. Budget primarily by trace volume, then refine with sampling and retention policy.

What’s the safest way to avoid cost surprises?

Set monthly usage limits, default retention to base, sample successful production traffic, keep 100% errors, and review running usage weekly.

Should I keep everything for 400 days?

Usually no. Extended retention is best for traces that carry long-term value: user feedback, curated dataset examples, and incident or regression investigations. Everything else can be base retention.

Can I estimate cost before launch?

Yes: estimate requests/day × days/month × tracing sampling rate. Then decide what fraction should be extended. Use the calculator above, and validate with a staging load test for one week.

Implementation tip: tracing only when you need it

Many teams trace 100% in dev/staging, but sample in production. During incidents and launches, they temporarily increase capture rate, then return to normal after stability is confirmed.

Example (policy idea, pseudo-config):
- dev: trace 100% (base retention)
- staging: trace 100% (base retention)
- prod:
  - errors: trace 100% (upgrade to extended if incident)
  - success: trace 10–20% (base retention)
  - user feedback traces: upgrade to extended

8) Disclaimer

This is an educational pricing guide. Rates, included allowances, and plan features can change. Your account may have promotions, startup discounts, enterprise terms, or configuration differences. Always verify current pricing and your workspace’s retention settings inside LangSmith and the official pricing pages.