LlmCli
llm is an open-source CLI tool and Python library for interacting with large language models from the command line. Created by Simon Willison, the project is available at github.com/simonw/llm, has approximately 11,900 GitHub stars, and was last updated in May 2026. It supports OpenAI, Anthropic Claude, Google Gemini, Meta Llama, and dozens of other providers via a plugin architecture.
Installation is a single command:
pip install llm # or: brew install llm / pipx install llm / uv tool install llm
llm keys set openai
llm "Ten fun names for a pet pelican"Prompts are piped naturally from stdin:
cat myfile.py | llm -s "Explain this code"
curl https://r.jina.ai/https://example.com | llm -s "Summarise this article"Providers beyond OpenAI are added via plugins: llm install llm-anthropic, llm install llm-gemini, llm install llm-ollama. Every prompt and response is logged automatically to SQLite and queryable via llm logs. The tool also supports embedding generation, multimodal input (images via -a), structured JSON output via schema definitions, and tool use (granting models the ability to execute functions).
Placed in Tools / Assess / inner.
llm occupies a different position from fabric (Tools/Assess/inner) — the two are complementary. fabric provides a curated library of reusable prompt patterns applied to piped input; llm provides the underlying infrastructure layer: model switching, API key management, automatic SQLite logging, embeddings, and a plugin ecosystem covering every major provider. They can be chained: cat doc.txt | fabric --pattern extract_wisdom | llm -s "Format this as a Markdown table".
The SQLite logging is a non-obvious differentiator. Every llm invocation is recorded with the full prompt, response, model, and timestamp — enabling cost tracking, prompt archaeology, and dataset construction from real usage without any additional instrumentation.
The plugin ecosystem is the widest of any CLI LLM tool. llm-ollama connects llm to the local Ollama runtime (also on this radar), making it a unified CLI across local and remote models. llm embed pairs with DuckDB Vector Search for generating and storing embeddings from the command line.
Inner position reflects 11,900 stars, active maintenance by a high-credibility author, and direct composability with Ollama, Jina AI Reader, DuckDB Vector Search, and fabric — all already on this radar.
Trial gate: llm used as the primary CLI for LLM interactions across at least two providers (e.g. OpenAI and Ollama), with SQLite logs reviewed after one week of use.
