Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/simonw/LLM/llms.txt

Use this file to discover all available pages before exploring further.

Plugins must be installed into the same virtual environment as LLM itself. The llm install command handles this automatically — it is a thin wrapper around pip install that always targets the correct environment regardless of how you installed LLM (Homebrew, pipx, a plain venv, etc.). Browse the Plugin Directory to find plugins worth installing.

​
Installing a plugin

1

Find a plugin

Check the Plugin Directory for a plugin that fits your needs. Each entry lists the package name you’ll pass to llm install.
2

Install from PyPI

Run llm install followed by the package name:
llm install llm-gpt4all
LLM fetches the package from PyPI and installs it into the right environment.
llm install requires pip to be available in the LLM environment. If you installed LLM with pipx or Homebrew, this is handled automatically. If you’re in a plain virtualenv, make sure pip is present before running the command.
3

Verify the new models or commands

After installation, confirm the plugin registered its contributions. For a model plugin, list available models:
llm models
Pass --options to see per-model configuration options:
llm models --options
Then run a prompt against a newly available model using -m:
llm -m orca-mini-3b-gguf2-q4_0 'What is the capital of France?'

​
Installing a plugin in editable mode

During plugin development you’ll want changes to your source tree to take effect immediately without reinstalling. Pass -e with the path to your local checkout: 
llm install -e /path/to/my-plugin
This installs the package in editable (development) mode, so edits to the source files are reflected instantly.

​
Useful install flags

llm install accepts several flags that are passed through to pip:
FlagDescription
-U / --upgradeUpgrade the package to the latest version on PyPI
--preInclude pre-release and development versions
--force-reinstallReinstall the package even if it is already up-to-date
--no-cache-dirDisable the pip download cache
# Upgrade an existing plugin to the latest release
llm install -U llm-anthropic

# Install a pre-release version
llm install --pre llm-anthropic

​
Listing installed plugins

Run llm plugins to see every plugin currently installed, the hooks each one registers, and its version: 
llm plugins

[
  {
    "name": "llm-anthropic",
    "hooks": [
      "register_models"
    ],
    "version": "0.11"
  },
  {
    "name": "llm-gguf",
    "hooks": [
      "register_commands",
      "register_models"
    ],
    "version": "0.1a0"
  },
  {
    "name": "llm-clip",
    "hooks": [
      "register_commands",
      "register_embedding_models"
    ],
    "version": "0.1"
  },
  {
    "name": "llm-cmd",
    "hooks": [
      "register_commands"
    ],
    "version": "0.2a0"
  },
  {
    "name": "llm-gemini",
    "hooks": [
      "register_embedding_models",
      "register_models"
    ],
    "version": "0.3"
  }
]

​
Uninstalling a plugin

Use llm uninstall to remove a plugin: 
llm uninstall llm-gpt4all
Add -y to skip the confirmation prompt: 
llm uninstall llm-gpt4all -y

​
Running with a subset of plugins

By default LLM loads every plugin installed in its environment. You can override this with the LLM_LOAD_PLUGINS environment variable — useful for debugging, performance testing, or temporarily disabling a misbehaving plugin. Disable all plugins: 
LLM_LOAD_PLUGINS='' llm prompt 'hello'
Load only specific plugins: 
LLM_LOAD_PLUGINS='llm-gpt4all,llm-cluster' llm prompt 'hello'
Verify which plugins are active by combining the variable with llm plugins: 
LLM_LOAD_PLUGINS='' llm plugins
LLM_LOAD_PLUGINS='' is also handy in CI pipelines or scripts where you want predictable behaviour without any user-installed plugins interfering.

​
Plugins that require extra dependencies

Some plugins — especially those that run models locally — have large optional dependencies (PyTorch, CUDA libraries, Metal frameworks, etc.) that are not installed by default. Check the plugin’s README for any extra install steps. A common pattern is to install the plugin first, then follow its documentation to pull in the model weights or framework: 
# 1. Install the plugin itself
llm install llm-mlx

# 2. Follow the plugin's README to download a model
llm mlx download-model mlx-community/Llama-3.2-3B-Instruct-4bit
Always refer to the individual plugin’s documentation for the exact setup instructions.

Build docs developers (and LLMs) love

Get started for freeTalk to us