A Swiss Army knife for LLMs that lives in your terminal

What it does

LLM is a command-line tool and Python library that lets you fire prompts at OpenAI, Anthropic, Google, Meta, and other providers without leaving your shell. It also logs every prompt and response to a local SQLite database by default, so your AI conversations become queryable history.

The interesting bit

The real depth is in the plugin architecture: you can bolt on new providers (including local models via Ollama) or capabilities like embeddings and structured data extraction, and the tool stays Unix-friendly—pipe a file into it, attach an image, or run an interactive chat without breaking your workflow.

Key highlights

• Talks to remote APIs and local models through a unified interface
• Built-in SQLite logging turns prompts into searchable records
• Supports multimodal attachments (images, audio, video) and structured output schemas
• Plugin ecosystem extends models, tools, and templates
• Templates and fragments help manage long-context prompts

Caveats

• The README flags a specific compatibility issue between the Homebrew distribution and PyTorch, so local model users may need to pick their install method carefully.

Verdict

Developers who live in the terminal and want to treat language models as composable Unix tools will feel right at home; if you prefer glossy web interfaces or only ever use one API, this probably isn’t your speed.