Skip to main content

Installation

Overview​

LangChain can be installed via pip or conda. For most use cases, pip is recommended as it will provide wider compatibility and more frequent updates. However, if you need to manage many other Python packages and environments, conda may be preferred.

The bare minimum installation via either method will include just the core LangChain library. For additional functionality, there are two extended installation options:

  • langchain[llms] - Installs dependencies for common LLM providers like OpenAI, Anthropic, Cohere etc. This allows integrating LangChain with popular LLMs out of the box.
  • langchain[all] - Installs dependencies for all integrations LangChain supports, including LLM providers, datastores, APIs, etc. Useful if you want to leverage LangChain's full capabilities, but will install many unneeded packages if you only use a subset of features.

Installation​

Pip​

To install the minimum requirements:

pip install langchain

To install with LLM provider dependencies:

pip install langchain[llms]

To install all integrations:

pip install 'langchain[all]'

Note: If using zsh, you'll need to quote the brackets to avoid issues.

Conda​

To install with conda:

conda install langchain -c conda-forge

The conda installation does not support installing optional dependencies like llms or all. Those would need to be installed separately after the fact.

Install from Source​

To install LangChain from source, clone the GitHub repo and run:

pip install -e .

This will install LangChain directly from the latest source code. Useful if you want to contribute to LangChain or need the very latest updates before they are released.

Next Steps​

With LangChain installed, you can now setup your environment and start building applications!