Getting Started

Prerequisites

  • Rust 1.94+ (if building from source)
  • A supported LLM provider account (OpenAI, or a local provider like Ollama / LM Studio)
  • Git (for session diffs and patch application)

Installation

From Source

git clone https://github.com/savfox-ai/savfox.git
cd savfox
cargo install --path crates/savfox-cli

Verify Installation

savfox --version

Authentication

Before using Savfox, authenticate with your LLM provider:

savfox login

This opens an interactive flow to configure your API credentials. Savfox stores credentials securely in your system keyring.

To use an open-source local provider instead:

savfox --oss login

To log out:

savfox logout

Your First Interactive Session

Simply run savfox with no arguments to launch the interactive terminal UI:

savfox

This opens a TUI (terminal user interface) where you can:

  • Type messages and chat with the AI agent
  • Review proposed file changes with diffs
  • Approve or reject commands and patches
  • Switch models and personalities

See Interactive Mode for details.

Your First Non-Interactive Execution

Use the exec subcommand (alias e) to run a one-shot task:

savfox exec "Add error handling to src/main.rs"

The agent will process your request, propose changes, and output results directly to the terminal. For JSON output (useful for scripting):

savfox exec --json "List all TODO comments in the codebase"

See CLI Reference for all available commands and options.

Quick Tips

  • Use --model or -m to specify a different LLM model:

    savfox -m gpt-4o exec "Explain this code"
  • Use --full-auto for low-friction automated execution:

    savfox --full-auto exec "Refactor the auth module"
  • Resume a previous session:

    savfox resume        # interactive picker
    savfox resume --last # resume the most recent session

Next Steps