LLM Rust — AI skill for Claude Code

LLM - Rust Edition

A fast, memory-safe reimplementation of the LLM CLI tool in Rust.

What this is

This is a Rust port of the Python `llm` CLI tool. It provides:

• Fast performance with Rust's zero-cost abstractions
• Memory safety without garbage collection
• Compatibility with existing workflows
• Support for multiple LLM providers

Features

Currently implemented

• Multi-provider support (OpenAI, Anthropic, OpenAI-compatible endpoints)
• Streaming responses with real-time token delivery
• Key management for API credentials
• Conversation tracking
• Model catalog with aliases
• Shell command generation via cmd subcommand
• Logging and usage analytics

In progress / next up

• Python plugin bridge for existing plugins
• Expanded third-party plugin command/model registration parity
• Broader provider coverage and model-catalog refresh workflows
• Continued tool-calling and structured-output parity hardening

Getting Started

Prerequisites

• Rust 1.70+ (install from rustup.rs)
• API keys for your chosen providers

Installation

git clone https://github.com/simonw/llm
cd llm-rust
cargo build --release
cargo run -- --help

Usage Examples

Basic prompts

# Ask a question
cargo run -- "What is the meaning of life?"

# Use a specific model
cargo run -- --model gpt-4 "Explain quantum computing"

# Disable streaming
cargo run -- --no-stream "Tell me a joke"

Key management

# Set your OpenAI key
cargo run -- keys set openai --value sk-...

# List stored keys
cargo run -- keys list

# Show key storage location
cargo run -- keys path

Command generation

# Generate shell commands
cargo run -- cmd "find all TypeScript files modified today"

# Auto-execute in non-interactive mode
LLM_CMD_AUTO_ACCEPT=1 cargo run -- cmd "show git status"

Model management

# List available models
cargo run -- models list

# Set default model
cargo run -- models default