LLM CLI

LLM is a CLI tool and Python library by Simon Willison that provides a unified interface for interacting with large language models. Run llm "explain this error", pipe output through models with cat error.log | llm "what went wrong?", or open interactive chats — every prompt and response is automatically logged to a local SQLite database for later review, search, and analysis.

A plugin architecture (100+ plugins) covers OpenAI, Anthropic Claude, Google Gemini, Meta Llama, and dozens more, via remote APIs or locally-run models through plugins like llm-ollama. It also includes embeddings generation and semantic search, reusable prompt templates, fragments, model aliases, tool use (function calling), and structured output via JSON schemas — making it a lightweight RAG and agent-building toolkit from the terminal. Limitations: it is a power-user tool with a learning curve around plugins and templates; and API costs still accrue per call like any other client.

pip install llm (or brew/pipx/uv), then llm keys set openai to configure provider keys.