LangChain Pricing Explained: Plans, Costs and What You Get

LangChain pricing explained: the framework is free, LangSmith is paid — Developer $0, Plus $39 a seat, Enterprise custom — usage in LSUs at $1 each.

By Galactic News Editorial · Published 2026-11-09T07:37:54+00:00 · Updated 2026-11-09T07:37:54+00:00

LangChain Pricing Explained: Plans, Costs and What You Get

LangChain pricing confuses people for one simple reason: the framework is free, and the bill comes from somewhere else. LangChain and LangGraph are open-source libraries that cost nothing to use, while LangSmith, the hosted platform for tracing, evaluation, and deployment, is where the money goes. This guide explains LangChain pricing plan by plan, including the usage unit that drives most of the bill, and works through what a small team actually pays. All prices below come from LangChain's own pricing page, read on 9 November 2026.

LangChain Pricing: What Is Free and What Is Paid

LangChain, LangGraph, and the LangSmith client libraries are free and open source under the MIT license. You can build agents, chains, and retrieval pipelines locally, call any model provider you like, and host everything yourself without paying LangChain anything. Your model API calls are billed by your provider, such as OpenAI or Anthropic, and your hosting is yours. What you pay LangChain for is LangSmith, the hosted platform that records every run of your agent as a trace, so you can replay failures, score answer quality, and deploy agents on managed infrastructure. LangSmith is optional: you can build with LangChain and never touch it. But once you ship an agent to real users, the tracing and evaluation features become hard to give up, and that is when the pricing below starts to matter.

The Three LangSmith Plans

LangSmith has three plans, and the headline prices are simple. Developer costs $0 per seat per month, covers one seat, and includes 5,000 base traces a month, then pay as you go. Plus costs $39 per seat per month, allows unlimited seats, and includes 10,000 base traces a month, with access to Deployment, Engine, and Fleet. Enterprise is custom priced, and it is the only tier with self-hosted and hybrid deployment, custom single sign-on, role-based access control, and a support SLA. A LangSmith for Startups offer provides discounted rates and up to $10,000 in credits for VC-backed companies. The plan choice is really a team-size choice: solo developers start on Developer, product teams move to Plus when they need a second seat or a deployment, and Enterprise is for organisations that cannot send traces to a vendor cloud.

LangChain Pricing and the LSU: The Unit That Drives the Bill

The part of LangChain pricing that surprises people is the usage meter. LangSmith measures most usage in LangChain Standard Units, or LSUs, and one LSU costs $1.00. Seats are a flat monthly charge, but everything beyond the included traces is metered in LSUs on top. A base trace is one recorded execution of your application, kept for 14 days, and extended traces are kept for 180 days. Deployment is metered by resources consumed: runtime compute at 0.0675 LSU per vCPU-hour, runtime memory at 0.0090 LSU per GiB-hour, database compute at 0.177 LSU per vCPU-hour, and database memory at 0.025 LSU per GiB-hour. Fleet, the product that builds agents from plain-language requests, includes 7 LSU a month on Developer and 37 LSU on Plus, then meters further usage. Sandboxes include 8 LSU a month, and Tuned Evaluators cost 0.015 LSU per evaluation run. The $39 seat price is only the floor: a busy agent can add far more in usage than the seat itself costs, so model your trace volume first.

A Worked Example: What a Small Deployment Costs

Take a small team on Plus with one seat and one small serverless deployment, which is included free on the plan. The seat costs $39 a month, and the included 10,000 base traces cover light usage. Now add the deployment's resource use. Suppose the deployment runs one virtual CPU for 100 hours in a month with 2 GiB of memory, and its database runs one virtual CPU for the same 100 hours with 1 GiB. Runtime compute costs 100 times 0.0675, which is 6.75 LSU, and runtime memory costs 200 GiB-hours times 0.0090, which is 1.80 LSU. Database compute costs 100 times 0.177, which is 17.70 LSU, and database memory costs 100 times 0.025, which is 2.50 LSU. The total is 28.75 LSU, or about $28.75 on top of the $39 seat, for a monthly total near $68 before trace overages. That is an estimate from the published metering rates, not a quote. LangChain provides a usage calculator on its pricing page, and you should run your own numbers through it before you commit. The key point is that the database is the expensive part of a deployment, at more than twice the runtime compute rate, so a deployment that holds a lot of state costs more than one that just answers questions.

What Each LangSmith Service Actually Does

LangSmith is five services under one roof, and the pricing page charges for them separately, so it helps to know what each one is. Observability and evaluation records traces, runs online and offline evals, manages datasets and prompt versions, and keeps an annotation queue for human feedback. Deployment hosts long-running agents, with one-click deploys, cron scheduling, and serverless or dedicated options. Fleet builds agents from everyday language, with prebuilt templates and remote MCP servers for tools. Engine watches production traces, clusters failures into named issues, and proposes fixes and new evals. Sandboxes are isolated environments for agent-generated code. The LLM Gateway sits between your agents and model providers, managing cost controls, rate limits, fallbacks, and redaction of sensitive data. You do not need all five services, and most teams start with observability alone, then add deployment when they ship.

LangChain Pricing: Common Traps to Avoid

Three traps catch new LangSmith users. The first is the seat count: a seat is a distinct user in your organisation, including invited users, so adding teammates to look at traces adds $39 per seat per month on Plus. The second is trace volume: the included 5,000 or 10,000 traces sound generous, but a chatty agent can generate thousands of traces a day, and overage is pay as you go with no published per-trace figure on the pricing page itself, so set up billing alerts early. The third is self-hosting: it is only available on Enterprise, so a team that needs data in its own cloud must pay the custom price. LangSmith states that it does not use your data to train models. Developer and Plus seats are billed monthly on the 1st, with traces billed monthly in arrears, while Enterprise is invoiced annually upfront.

Five Steps to Estimate Your LangSmith Bill

1. Step 1: Count your seats. Multiply $39 by each Plus seat, or stay on Developer for one free seat.
2. Step 2: Estimate monthly traces. Compare the number against the 5,000 or 10,000 included traces.
3. Step 3: Size your deployment. Work out vCPU-hours and GiB-hours from the metering rates above.
4. Step 4: Add Fleet, Engine, and Sandboxes. Check the included LSU allowances before counting extra usage.
5. Step 5: Run the usage calculator. Put your numbers into LangChain's calculator and set a billing alert.

Bottom Line

LangChain pricing is free for the framework and paid for the platform, and the platform bill has two parts: seats and LSUs. A solo developer pays nothing on the Developer plan with 5,000 traces a month. A small team on Plus pays $39 per seat per month plus usage at $1 per LSU, and a realistic small deployment adds roughly $30 a month in metering. Enterprise is custom and is the only way to self-host. Before you commit, estimate your trace volume and run the usage calculator on LangChain's pricing page. If your agent is still a prototype, stay on Developer and pay nothing until the traces tell you it is time to move.