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.

LLM is a CLI tool and Python library that lets you interact with large language models from the command line or from Python code. With a single, consistent interface you can run prompts against OpenAI’s GPT models, Anthropic’s Claude, Google’s Gemini, Meta’s Llama, and dozens of other models — both hosted APIs and models running entirely on your own machine. Whether you want to quickly summarize a file, extract structured data from text, build an automated pipeline, or just explore what the latest models can do, LLM gives you a fast and composable way to do it without writing boilerplate code.
LLM ships with OpenAI support built in. The default model is gpt-4o-mini — fast and inexpensive for everyday use. You can change this at any time with llm models default <model-id>.

​
Key Features

Command-Line Prompts

Run single prompts, pipe files through system prompts, or attach images and documents — all from your terminal. Works seamlessly with standard Unix pipes and shell scripting.

Interactive Chat

Start a persistent, multi-turn conversation with any model using llm chat. Supports multi-line input, editor integration, and fragment injection during the session.

SQLite Logging

Every prompt and response is automatically logged to a local SQLite database. Search, filter, and replay your entire history — or browse it visually with Datasette.

Plugin Ecosystem

Extend LLM with plugins that add new models, tools, embedding providers, template loaders, and CLI commands. Install with a single llm install command.

Python API

Use LLM as a Python library for scripting, automation, and integrations. The same model plugins that work on the CLI work seamlessly in Python code.

Multi-Modal Support

Attach images, audio, video, and documents to your prompts. Extract text from scanned PDFs, describe screenshots, and more — with any model that supports attachments.

Embeddings

Generate and store vector embeddings from any text using embedding model plugins. Run semantic similarity searches directly from the CLI or Python.

Structured Output

Extract structured JSON from unstructured content using schemas. Define a schema inline or as a reusable template, and LLM will return validated, typed data.

​
The Plugin Ecosystem

LLM’s plugin system is the heart of its extensibility. Plugins are standard Python packages distributed via PyPI and installed with llm install <plugin-name>. A few of the most popular plugins:
PluginAdds
llm-anthropicAnthropic Claude models (claude-4-opus, claude-3.5-haiku, …)
llm-geminiGoogle Gemini models (gemini-2.0-flash, gemini-1.5-pro, …)
llm-ollamaAny model served locally via Ollama
llm-gpt4all17+ local models via GPT4All (no API key required)
llm-sentence-transformersLocal embedding models from Sentence Transformers
llm-clipCLIP image embeddings for visual similarity search
Browse the full plugin directory in the official documentation.

​
What Makes LLM Distinctive

One tool, every model. Rather than learning a separate SDK or CLI for OpenAI, Anthropic, Gemini, and local models, LLM provides a single, uniform interface. Switch models with -m <model-id> and everything else stays the same. Composable by design. LLM embraces Unix philosophy. Pipe text in, get text out. Chain it with grep, jq, cat, curl, and any other tool in your shell. Use it in shell scripts, Makefiles, and CI pipelines. Your data stays yours. All logs are stored in a local SQLite database on your own machine. No data is sent anywhere except to the model provider you choose. Extensible at every layer. Plugins can add new models, new embedding providers, new CLI subcommands, new template loaders, and new fragment sources. The plugin API is stable and well-documented.

Build docs developers (and LLMs) love

Get started for freeTalk to us