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.

By Mackenzie Bowes · mackenzie@mackenziebowes.com · @_mackenziebowes on X

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]

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]

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

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

  1. LangChain, Agents (Docs by LangChain, Python), read 2026-11-06, undated. docs.langchain.com/oss/python/langchain/agents
  2. LangChain, LangChain overview (Docs by LangChain, Python), read 2026-11-05, undated. docs.langchain.com/oss/python/langchain/overview
  3. langchain-ai, langchain repository, LICENSE file (MIT License), read 2026-11-06. github.com/langchain-ai/langchain/blob/master/LICENSE
  4. 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.