CLI Usage

The Unfault CLI is the primary interface for analyzing your codebase and getting production-readiness feedback.

Installation

The Unfault CLI is available for macOS, Linux, and Windows as an executable binary. You can install it using one of the following methods.

From releases

Download the right release for your platform from GitHub Releases and add it to your PATH.

curl -sSL https://unfault.dev/get | bash

From crates.io

cargo install unfault

From source

git clone https://github.com/unfault/cli
cd cli
cargo build --release

Commands

The unfault cli has a set of commands that you can use to interact with the Unfault platform.

unfault login

Authenticate using secure device flow. No API keys in your terminal history.

unfault login
# Visit https://app.unfault.dev/auth/device/ and enter the displayed code

Note

Your authentication token is stored securely and used for all subsequent commands. See the Configuration section for file locations on each platform.

unfault review

Analyze your codebase for production-readiness issues.

# Basic output (grouped by severity)
unfault review

# Full details with suggested fixes
unfault review --output full

# JSON for integration with other tools
unfault review --output json

# Focus on specific dimensions
unfault review --dimension stability --dimension performance

Output Modes:

Mode	Description
basic	Grouped by severity, rule counts (default)
concise	Summary statistics only
full	Detailed findings with diffs
json	Machine-readable output

unfault ask

Query your project’s health using natural language.

# Ask about your codebase
unfault ask "What are my main stability concerns?"

# Scope to a specific workspace
unfault ask "Show recent issues" --workspace wks_abc123

# Get raw context without AI synthesis
unfault ask "Performance problems" --no-llm

Note

The ask command requires you to have run unfault review at least once.

Configure an LLM for AI-powered answers:

OpenAI

unfault config llm openai --model gpt-5.1

Anthropic

unfault config llm anthropic --model claude-sonnet-4-5

Ollama

unfault config llm ollama --model llama3.2

Tip

When asking questions, you can use the --no-llm flag to get raw context without AI synthesis.

unfault status

Check authentication and connectivity.

unfault status

unfault config

Manage CLI configuration.

# Show current config
unfault config show

# Configure LLM provider
unfault config llm openai --model gpt-4o

# View LLM settings
unfault config llm show

# Remove LLM configuration
unfault config llm remove

Exit Codes

The CLI uses standard exit codes for CI/CD integration:

Code	Meaning	Action
0	Success, no issues	✅ Proceed
1	General error	🔍 Check logs
2	Configuration error	Run unfault login
3	Authentication failed	Re-authenticate
4	Network error	Check connectivity
5	Findings detected	🚨 Review issues
6	Invalid input	Check arguments
7	Service unavailable	Retry later
8	Session error	Retry analysis
10	Subscription required	Upgrade plan

Configuration

The CLI configuration file location depends on your operating system:

Platform	Location
Linux	~/.config/unfault/config.json or $XDG_CONFIG_HOME/unfault/config.json
macOS	~/.config/unfault/config.json or $XDG_CONFIG_HOME/unfault/config.json
Windows	%USERPROFILE%\.config\unfault\config.json

Note

All platforms use the same .config directory pattern for consistency, rather than platform-specific locations like ~/Library/Application Support (macOS) or %APPDATA% (Windows).

Example configuration:

{
  "api_key": "uf_live_...",
  "base_url": "https://api.unfault.dev",
  "llm": {
    "provider": "openai",
    "model": "gpt-4",
    "api_key": "sk-..."
  }
}

Environment Variables

Variable	Description
UNFAULT_BASE_URL	Override API endpoint
OPENAI_API_KEY	OpenAI API key (for ask command)
ANTHROPIC_API_KEY	Anthropic API key (for ask command)

Using with AI Agents

Tip

Unfault works well with AI coding assistants like Claude or ChatGPT. Add this to your AGENTS.md or system prompt:

# Code Quality

Run `unfault review` before committing to check for production issues.
Use `unfault review --output full` to see suggested patches.

The CLI returns structured output that AI agents can parse and act on.