AI Integration

Get AI assistance without leaving the terminal. mine ai supports Claude, OpenAI, Gemini, and OpenRouter for code review, commit message generation, and general questions — with keys stored securely in the vault.

Key Capabilities

• Multi-provider — Claude, OpenAI, Gemini, and OpenRouter supported out of the box
• Code review — send staged git diffs for AI review in one command
• Commit messages — generate conventional commit messages from staged changes
• Quick questions — ask coding questions directly from the terminal
• Styled markdown output — responses rendered as formatted markdown in interactive terminals (headings, code blocks, lists, emphasis)
• Secure key storage — API keys stored in the encrypted vault, or use environment variables
• System instructions — customize AI behavior globally or per-subcommand via config or --system flag

Quick Example

# Configure a provider (key stored in vault)
mine ai config --provider claude --key sk-ant-...

# Ask a question — response rendered as styled markdown in TTY
mine ai ask "What's the difference between a mutex and a semaphore?"

# Review staged changes — review rendered as styled markdown in TTY
git add . && mine ai review

# Generate a commit message
git add . && mine ai commit

# Force plain markdown output (useful for piping to files or other tools)
mine ai ask "Explain goroutines" --raw > answer.md
mine ai review --raw | less

# Override AI behavior for a single invocation
mine ai review --system "Focus only on security issues."
mine ai ask "Explain Go channels" --system "You are a Go expert. Be concise."

How It Works

Configure a provider once with mine ai config and the API key is stored encrypted in the vault. From then on, AI commands just work. mine ai review sends your staged diff to the configured provider and returns feedback. mine ai commit analyzes staged changes and suggests a conventional commit message, optionally running git commit for you.

For zero-config setups, just set the standard environment variable (ANTHROPIC_API_KEY, OPENAI_API_KEY, etc.) and mine detects it automatically. Environment variables take precedence over vault-stored keys.

System instructions let you tailor how the AI responds. Pass --system "<text>" on any ai subcommand to override behavior for that invocation, or set persistent defaults with mine config set. A four-level precedence chain (flag → subcommand config → global config → built-in default) gives you fine-grained control without breaking existing workflows.

Use mine ai models to see all available providers, their suggested models, and whether they’re configured.

Markdown Rendering

In interactive terminals (TTY), mine ai ask and mine ai review automatically render AI responses as styled markdown — headings, code blocks, lists, bold/italic text are all formatted for readability using glamour.

When piping output or running in a non-TTY context, raw markdown is emitted automatically — no flags needed. Use --raw to force plain output even in a TTY (useful when saving to a file or chaining with other tools):

# TTY: styled output (default)
mine ai ask "How does Go garbage collection work?"

# Non-TTY: raw markdown automatically
mine ai ask "How does Go garbage collection work?" | cat

# Force raw regardless of context
mine ai ask "How does Go garbage collection work?" --raw > answer.md

Learn More

See the command reference for provider setup, model selection, and all subcommands.