LLM CLI Cheatsheet

Q: How do I install and configure llm?

Run `pip install llm`. Then set an API key: `llm keys set openai` (enter your key when prompted). Install model plugins as needed: `pip install llm-claude-3` for Claude, `pip install llm-gemini` for Gemini, or use Ollama for local models.

Q: Can I use local models with llm?

Yes, via the Ollama plugin: `pip install llm-ollama`, then run `llm -m ollama 'prompt'`. Or use GPT4All: `pip install llm-gpt4all`, then `llm -m gpt4all 'prompt'`. Local models work completely offline.

Q: How do I use llm for embeddings and RAG?

Generate embeddings: `llm embed -m text-embedding-3-small -i 'text'`. Batch embed documents: `llm embed-multi docs -m text-embedding-3-small --files ./docs/ '*.md'`. Then search: `llm similar docs -c 'query text'`. Store embeddings in a SQLite database for persistence.

Q: What are prompt templates?

Templates are reusable prompts with variables. Create one with `llm templates edit my-template`, write something like 'Summarize this: {{input}}'. Then use it: `llm -t my-template 'long text...'`. Templates live in `~/.llm/templates/` as YAML files.

`llm` is a lightweight command-line LLM tool supporting OpenAI, Claude, Gemini, Ollama, GPT4All, and more via plugins. Features include prompt templates, interactive chat, embedding generation, and semantic search — the most convenient way to interact with LLMs from the terminal.

Updated: 2026-07-20·24 commands

Quick Start

``bash # Install pip install llm

# Set an API key llm keys set openai

# Ask a question llm 'What is the capital of Japan?'

# Use a different model llm -m claude-3-haiku-20240307 'Write a haiku'

# Chat interactively llm chat

# Continue the conversation llm -c 'Tell me more'`

`Supported Models`

Out of the box: OpenAI (GPT-4o, GPT-4o-mini). Via plugins: Claude (llm-claude-3), Gemini (llm-gemini), Ollama (llm-ollama, for local models), GPT4All (llm-gpt4all, fully offline), and many community plugins.

`Chat & Conversation`

llm chat starts an interactive session. Use -c to continue any previous conversation. Use llm logs to browse history. Conversations include system prompts, user messages, and model responses in a persistent SQLite database.

`Prompt Templates`

Create reusable prompt templates with llm templates edit . Templates use Jinja-style {{variable}} syntax. Use them with llm -t 'value'. Pre-installed templates include various summarization and extraction patterns.

`Embeddings & Semantic Search`

llm includes a complete embedding workflow: llm embed for single texts, llm embed-multi for batch embedding documents, and llm similar for semantic search. Embeddings are stored in SQLite and can be exported for use in other tools.

`Plugin System`

llm's power comes from its plugin ecosystem. Install plugins with pip install llm-. List installed plugins with llm plugins`. Community plugins add models, embedding functions, template loaders, and more.

Startup & Modes(20)

Command	Level	Example
`llm --help` Show llm CLI help	Basic	`llm --help`
`llm 'prompt'` Query LLM directly with prompt argument (core usage)	Basic	`llm 'What is the capital of France?'`
`llm -m` Specify which model to use	Basic	`llm -m gpt-4o 'Explain quantum computing'`
`llm -m claude-3-opus-20240307` Use Claude Opus model	Basic	`llm -m claude-3-opus-20240307 'Write a poem'`
`llm -m gpt4all` Use local GPT4All model (offline)	Intermediate	`llm -m gpt4all 'What is Rust used for?'`
`llm -m ollama` Use Ollama-managed local model	Intermediate	`llm -m ollama 'Tell me a joke'`
`llm -s` Set system prompt	Basic	`llm -s 'You are a pirate' 'How do I make coffee?'`
`llm logs` View recent conversation history	Basic	`llm logs`
`llm logs -n` Show last N conversation records	Basic	`llm logs -n 20`
`llm -c` Continue last conversation (appends to context)	Intermediate	`llm -c 'Tell me more about that'`
`llm chat` Start interactive chat mode	Basic	`llm chat`
`llm chat -m` Specify model in interactive mode	Basic	`llm chat -m claude-3-haiku-20240307`
`llm keys` List configured API keys	Basic	`llm keys`
`llm keys set` Set an API key	Basic	`llm keys set openai`
`llm keys get` Get stored API key value	Basic	`llm keys get openai`
`pip install llm` Install llm CLI tool	Basic	`pip install llm`
`pip install llm-claude-3` Install Claude plugin	Basic	`pip install llm-claude-3`
`llm plugins` List installed plugins	Basic	`llm plugins`
`llm templates` List prompt templates	Intermediate	`llm templates`
`llm templates edit` Edit a prompt template	Intermediate	`llm templates edit my-template`

ai-embedding(4)

Command	Level	Example
`llm embed` Generate text embeddings	Intermediate	`llm embed -m text-embedding-3-small -i 'Hello world'`
`llm embed-multi` Batch generate embeddings (from file)	Expert	`llm embed-multi articles -m text-embedding-3-small -d articles.jsonl`
`llm embed-multi --files` Batch embed from directory files	Expert	`llm embed-multi docs -m text-embedding-3-small --files ./docs/ '*.md'`
`llm similar` Query semantically similar text	Expert	`llm similar articles -c 'machine learning'`

FAQ

How do I install and configure llm?

Can I use local models with llm?

How do I use llm for embeddings and RAG?

What are prompt templates?

This cheatsheet is compiled from official tool documentation. Last updated: 2026-07-20.