Building an AI agent with LangChain
LangChain's docs define an agent as a model calling tools in a loop and offer create_agent, LangGraph and Deep Agents. The library is MIT-licensed; its tracing service, LangSmith, is free for one seat to 5,000 traces a month, then 39 dollars a seat. Read on 2026-11-06.
Contents
LangChain is a Python library, with companion products, for building agents on language models. This page reads LangChain's own documentation, licence file and price list on 2026-11-06 to answer three questions people search: what a LangChain agent is, how to start one, and what it costs. It states what the vendor says, not what a test showed, and covers the Python documentation only. How long a build takes, as agencies quote it and as 47 enterprise rollouts measured it, is on How long does it take to build an AI agent?.
What an agent is in LangChain's words
The agents page says: "An agent is a model calling tools in a loop until a given task is complete." It adds: "A harness is everything around that loop: the prompt, the tools, and any middleware that shapes the model's behavior", and the formula "Agent = Model + Harness."[1] Its function is create_agent, "a highly configurable harness": the minimum is a model name and a list of tools, for example create_agent(model="google_genai:gemini-3.6-flash", tools=tools), and system prompts, middleware, structured output, memory, streaming and human-in-the-loop checks are the listed extras.[1] The model-provider tabs on that page list Google, OpenAI, Anthropic, OpenRouter, Fireworks, Baseten and Ollama.[1]
Four products, four jobs
The overview page separates the products:[2]
- LangChain (
create_agent): "a highly customizable harness, easily tailored to your use case and data." - LangGraph: "our low-level orchestration framework, for advanced needs combining deterministic and agentic workflows."
- Deep Agents: a "batteries-included harness (context compression, virtual filesystem, subagents)", offered as a managed service or as a software kit you host yourself.
- LangSmith: "to trace, debug, and evaluate agents built with any of these frameworks."
The agents page adds that "Managed Deep Agents builds on create_agent", and the overview says LangChain's agents "are built on top of LangGraph."[1][2] The docs offer these as the choices; they do not test which one suits which job.
Licence
The repository's LICENSE file is the MIT License.[3] Another Python agent framework under the MIT licence, CrewAI, is read on CrewAI: human input, tutorial, licence, price.
Tracing and its price
The overview tells a developer: "Set LANGSMITH_TRACING=true and your API key to get started."[2] Unlike the OpenAI Agents SDK, which says tracing is enabled by default (see OpenAI Agents SDK: start, MCP and tracing), LangChain's page presents tracing as a setting you turn on. Whether any LangChain package sends data without that setting is not shown by the pages read.
LangSmith's price list shows:[4]
- Developer: "$0 / seat per month", one seat, "Up to 5k base traces / mo, then pay-as-you-go".
- Plus: "$39 / seat per month", "Up to 10k base traces / mo".
- Enterprise: by contact.
- Units: "LangChain Standard Units. $1.00 / LSU", and the usage calculator shows "0.005 LSU / additional trace" beyond the included 10,000 traces on Plus. At one LSU to the dollar, that is $0.005 a trace, or $5 for 1,000 extra traces (this page's arithmetic).
A discount for venture-backed startups is offered, "Get up to $10,000 in credits."[4] The price list shows tracing and other services priced separately; the cost of the models an agent calls is not on it.
Not yet proven
- That a LangChain agent works reliably in production. The sources are the vendor's documentation and price list; no test result is cited, and no run was made.
- Which of the four products suits a small business. The docs list the options without a comparison on cost or effort.
- What the model calls cost. LangSmith's list covers tracing and related services only.
- Whether tracing stays off without the setting. The docs say how to turn it on; they do not state the default for every package.
- That these are stable interfaces. The pages show recent model names and products such as Managed Deep Agents, which a later docs version may change.
If you build agents with LangChain and have measured what tracing and the model calls cost at your volume, I would like to compare notes: mackenzie@mackenziebowes.com.
Sources
- LangChain, Agents (Docs by LangChain, Python), read 2026-11-06, undated. docs.
langchain. com/ oss/ python/ langchain/ agents - LangChain, LangChain overview (Docs by LangChain, Python), read 2026-11-05, undated. docs.
langchain. com/ oss/ python/ langchain/ overview - langchain-ai, langchain repository, LICENSE file (MIT License), read 2026-11-06. github.
com/ langchain- ai/ langchain/ blob/ master/ LICENSE - LangChain, LangSmith Plans and Pricing, read 2026-11-06, undated. www.
langchain. com/ pricing
Terms on this page
- Agentic workflow: A multistep process in which AI agents decide what to do next, instead of following fixed rules.
- AI agent: A system where a language model decides for itself how to do a task and which tools to use.
- Human in the loop: A person takes part in an automated system's work, supervision or decisions.
- LangChain: A software toolkit from LangChain for building AI agents and applications on language models.
- Model Context Protocol: An open standard for connecting AI applications to outside tools and data.